Hub Card Data
Collection
2 items
•
Updated
•
4
Datasets:
Dataset Viewer
modelId
stringlengths 5
139
| author
stringlengths 2
42
| last_modified
timestamp[us, tz=UTC]date 2020-02-15 11:33:14
2025-06-07 06:27:44
| downloads
int64 0
223M
| likes
int64 0
11.7k
| library_name
stringclasses 469
values | tags
sequencelengths 1
4.05k
| pipeline_tag
stringclasses 54
values | createdAt
timestamp[us, tz=UTC]date 2022-03-02 23:29:04
2025-06-07 06:26:26
| card
stringlengths 11
1.01M
|
|---|---|---|---|---|---|---|---|---|---|
mradermacher/QoQ-Med-VL-32B-i1-GGUF | mradermacher | 2025-06-07T05:00:06 | 0 | 0 | transformers | [
"transformers",
"gguf",
"en",
"base_model:ddvd233/QoQ-Med-VL-32B",
"base_model:quantized:ddvd233/QoQ-Med-VL-32B",
"license:mit",
"endpoints_compatible",
"region:us",
"imatrix",
"conversational"
] | null | 2025-06-07T01:50:50 | ---
base_model: ddvd233/QoQ-Med-VL-32B
language:
- en
library_name: transformers
license: mit
quantized_by: mradermacher
---
## About
<!-- ### quantize_version: 2 -->
<!-- ### output_tensor_quantised: 1 -->
<!-- ### convert_type: hf -->
<!-- ### vocab_type: -->
<!-- ### tags: nicoboss -->
weighted/imatrix quants of https://huggingface.co/ddvd233/QoQ-Med-VL-32B
<!-- provided-files -->
static quants are available at https://huggingface.co/mradermacher/QoQ-Med-VL-32B-GGUF
## Usage
If you are unsure how to use GGUF files, refer to one of [TheBloke's
READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
more details, including on how to concatenate multi-part files.
## Provided Quants
(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
| Link | Type | Size/GB | Notes |
|:-----|:-----|--------:|:------|
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ1_S.gguf) | i1-IQ1_S | 7.4 | for the desperate |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ1_M.gguf) | i1-IQ1_M | 8.0 | mostly desperate |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 9.1 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ2_XS.gguf) | i1-IQ2_XS | 10.1 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ2_S.gguf) | i1-IQ2_S | 10.5 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ2_M.gguf) | i1-IQ2_M | 11.4 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q2_K_S.gguf) | i1-Q2_K_S | 11.6 | very low quality |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q2_K.gguf) | i1-Q2_K | 12.4 | IQ3_XXS probably better |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 12.9 | lower quality |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ3_XS.gguf) | i1-IQ3_XS | 13.8 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q3_K_S.gguf) | i1-Q3_K_S | 14.5 | IQ3_XS probably better |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ3_S.gguf) | i1-IQ3_S | 14.5 | beats Q3_K* |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ3_M.gguf) | i1-IQ3_M | 14.9 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q3_K_M.gguf) | i1-Q3_K_M | 16.0 | IQ3_S probably better |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q3_K_L.gguf) | i1-Q3_K_L | 17.3 | IQ3_M probably better |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-IQ4_XS.gguf) | i1-IQ4_XS | 17.8 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q4_0.gguf) | i1-Q4_0 | 18.8 | fast, low quality |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q4_K_S.gguf) | i1-Q4_K_S | 18.9 | optimal size/speed/quality |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q4_K_M.gguf) | i1-Q4_K_M | 20.0 | fast, recommended |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q4_1.gguf) | i1-Q4_1 | 20.7 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q5_K_S.gguf) | i1-Q5_K_S | 22.7 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q5_K_M.gguf) | i1-Q5_K_M | 23.4 | |
| [GGUF](https://huggingface.co/mradermacher/QoQ-Med-VL-32B-i1-GGUF/resolve/main/QoQ-Med-VL-32B.i1-Q6_K.gguf) | i1-Q6_K | 27.0 | practically like static Q6_K |
Here is a handy graph by ikawrakow comparing some lower-quality quant
types (lower is better):
And here are Artefact2's thoughts on the matter:
https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
## FAQ / Model Request
See https://huggingface.co/mradermacher/model_requests for some answers to
questions you might have and/or if you want some other model quantized.
## Thanks
I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
me use its servers and providing upgrades to my workstation to enable
this work in my free time. Additional thanks to [@nicoboss](https://huggingface.co/nicoboss) for giving me access to his private supercomputer, enabling me to provide many more imatrix quants, at much higher quality, than I would otherwise be able to.
<!-- end -->
|
kowndinya23/ultrafeedback_binarized-alpaca-llama-3-1b-2-epochs-alpha-0-beta-0.2-2-epochs | kowndinya23 | 2025-06-07T04:50:05 | 0 | 0 | transformers | [
"transformers",
"safetensors",
"llama",
"text-generation",
"generated_from_trainer",
"trl",
"dpo",
"conversational",
"dataset:trl-lib/ultrafeedback_binarized",
"arxiv:2305.18290",
"base_model:kowndinya23/alpaca-cleaned-llama-3-1b-2-epochs-alpha-0-beta-0.2",
"base_model:finetune:kowndinya23/alpaca-cleaned-llama-3-1b-2-epochs-alpha-0-beta-0.2",
"autotrain_compatible",
"text-generation-inference",
"endpoints_compatible",
"region:us"
] | text-generation | 2025-06-07T03:54:28 | ---
base_model: kowndinya23/alpaca-cleaned-llama-3-1b-2-epochs-alpha-0-beta-0.2
datasets: trl-lib/ultrafeedback_binarized
library_name: transformers
model_name: ultrafeedback_binarized-alpaca-llama-3-1b-2-epochs-alpha-0-beta-0.2-2-epochs
tags:
- generated_from_trainer
- trl
- dpo
licence: license
---
# Model Card for ultrafeedback_binarized-alpaca-llama-3-1b-2-epochs-alpha-0-beta-0.2-2-epochs
This model is a fine-tuned version of [kowndinya23/alpaca-cleaned-llama-3-1b-2-epochs-alpha-0-beta-0.2](https://huggingface.co/kowndinya23/alpaca-cleaned-llama-3-1b-2-epochs-alpha-0-beta-0.2) on the [trl-lib/ultrafeedback_binarized](https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized) dataset.
It has been trained using [TRL](https://github.com/huggingface/trl).
## Quick start
```python
from transformers import pipeline
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="kowndinya23/ultrafeedback_binarized-alpaca-llama-3-1b-2-epochs-alpha-0-beta-0.2-2-epochs", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])
```
## Training procedure
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://adobesensei.wandb.io/hrenduchinta/huggingface/runs/2utr7hi9)
This model was trained with DPO, a method introduced in [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290).
### Framework versions
- TRL: 0.18.1
- Transformers: 4.52.4
- Pytorch: 2.7.0
- Datasets: 3.6.0
- Tokenizers: 0.21.1
## Citations
Cite DPO as:
```bibtex
@inproceedings{rafailov2023direct,
title = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
author = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
year = 2023,
booktitle = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
url = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
editor = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
}
```
Cite TRL as:
```bibtex
@misc{vonwerra2022trl,
title = {{TRL: Transformer Reinforcement Learning}},
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
year = 2020,
journal = {GitHub repository},
publisher = {GitHub},
howpublished = {\url{https://github.com/huggingface/trl}}
}
``` |
minhxle/truesight-ft-job-7c73b7bb-e15b-467f-b887-567bdee795e3 | minhxle | 2025-06-07T02:26:37 | 0 | 0 | transformers | [
"transformers",
"safetensors",
"text-generation-inference",
"unsloth",
"qwen2",
"trl",
"en",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | null | 2025-06-07T02:26:24 | ---
base_model: unsloth/qwen2.5-7b-instruct-unsloth-bnb-4bit
tags:
- text-generation-inference
- transformers
- unsloth
- qwen2
- trl
license: apache-2.0
language:
- en
---
# Uploaded model
- **Developed by:** minhxle
- **License:** apache-2.0
- **Finetuned from model :** unsloth/qwen2.5-7b-instruct-unsloth-bnb-4bit
This qwen2 model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|
echo840/MonkeyOCR | echo840 | 2025-06-07T01:52:20 | 0 | 2 | transformers | [
"transformers",
"safetensors",
"visual-document-retrieval",
"arxiv:2506.05218",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | visual-document-retrieval | 2025-06-05T03:09:54 | ---
license: apache-2.0
pipeline_tag: visual-document-retrieval
library_name: transformers
---
<div align="center" xmlns="http://www.w3.org/1999/html">
<h1 align="center">
MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm
</h1>
[](https://arxiv.org/abs/2506.05218)
[](https://huggingface.co/echo840/MonkeyOCR)
[](https://github.com/Yuliang-Liu/MonkeyOCR/issues?q=is%3Aopen+is%3Aissue)
[](https://github.com/Yuliang-Liu/MonkeyOCR/issues?q=is%3Aissue+is%3Aclosed)
[](https://github.com/Yuliang-Liu/MonkeyOCR)
</div>
> **MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm**<br>
> Zhang Li, Yuliang Liu, Qiang Liu, Zhiyin Ma, Ziyang Zhang, Shuo Zhang, Zidun Guo, Jiarui Zhang, Xinyu Wang, Xiang Bai <br>
[](https://arxiv.org/abs/2506.05218)
[](README.md)
[](https://huggingface.co/echo840/MonkeyOCR)
[](http://vlrlabmonkey.xyz:7685/)
## Introduction
MonkeyOCR adopts a Structure-Recognition-Relation (SRR) triplet paradigm, which simplifies the multi-tool pipeline of modular approaches while avoiding the inefficiency of using large multimodal models for full-page document processing.
1. Compared with the pipeline-based method MinerU, our approach achieves an average improvement of 5.1% across nine types of Chinese and English documents, including a 15.0% gain on formulas and an 8.6% gain on tables.
2. Compared to end-to-end models, our 3B-parameter model achieves the best average performance on English documents, outperforming models such as Gemini 2.5 Pro and Qwen2.5 VL-72B.
3. For multi-page document parsing, our method reaches a processing speed of 0.84 pages per second, surpassing MinerU (0.65) and Qwen2.5 VL-7B (0.12).
<img src="https://v1.ax1x.com/2025/06/05/7jQ3cm.png" alt="7jQ3cm.png" border="0" />
## News
* ```2025.06.05 ``` 🚀 We release MonkeyOCR, which supports the parsing of various types of Chinese and English documents.
## Quick Start
### 1. Install MonkeyOCR
```bash
conda create -n MonkeyOCR python=3.10
conda activate MonkeyOCR
git clone https://github.com/Yuliang-Liu/MonkeyOCR.git
cd MonkeyOCR
# Install pytorch, see https://pytorch.org/get-started/previous-versions/ for your cuda version
pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --index-url https://download.pytorch.org/whl/cu124
pip install .
```
### 2. Download Model Weights
```python
pip install huggingface_hub
python download_model.py
```
### 3. Inference
```bash
# Make sure in MonkeyOCR directory
python parse.py path/to/your.pdf
# Specify MonkeyChat path and model configs path
python parse.py path/to/your.pdf -m model_weight/Recognition -c config.yaml
```
### 4. Gradio demo
```bash
# Prepare your env for gradio
pip install gradio==5.23.3
pip install pdf2image==1.17.0
```
```bash
# Start demo
python demo/demo_gradio.py
```
Using the [LMDeploy](https://github.com/InternLM/lmdeploy), our model can run efficiently on an NVIDIA 3090 GPU.
## Benchmark Results
Here are the evaluation results of our model on OmniDocBench. MonkeyOCR-3B uses DocLayoutYOLO as the structure detection model, while MonkeyOCR-3B* uses our trained structure detection model with improved Chinese performance.
### 1. The end-to-end evaluation results of different tasks.
<table style="width:100%; border-collapse:collapse; text-align:center;" border="0">
<thead>
<tr>
<th rowspan="2">Model Type</th>
<th rowspan="2">Methods</th>
<th colspan="2">Overall Edit↓</th>
<th colspan="2">Text Edit↓</th>
<th colspan="2">Formula Edit↓</th>
<th colspan="2">Formula CDM↑</th>
<th colspan="2">Table TEDS↑</th>
<th colspan="2">Table Edit↓</th>
<th colspan="2">Read Order Edit↓</th>
</tr>
<tr>
<th>EN</th>
<th>ZH</th>
<th>EN</th>
<th>ZH</th>
<th>EN</th>
<th>ZH</th>
<th>EN</th>
<th>ZH</th>
<th>EN</th>
<th>ZH</th>
<th>EN</th>
<th>ZH</th>
<th>EN</th>
<th>ZH</th>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="7">Pipeline Tools</td>
<td>MinerU</td>
<td>0.150</td>
<td>0.357</td>
<td>0.061</td>
<td>0.215</td>
<td>0.278</td>
<td>0.577</td>
<td>57.3</td>
<td>42.9</td>
<td>78.6</td>
<td>62.1</td>
<td>0.180</td>
<td>0.344</td>
<td><strong>0.079</strong></td>
<td>0.292</td>
</tr>
<tr>
<td>Marker</td>
<td>0.336</td>
<td>0.556</td>
<td>0.080</td>
<td>0.315</td>
<td>0.530</td>
<td>0.883</td>
<td>17.6</td>
<td>11.7</td>
<td>67.6</td>
<td>49.2</td>
<td>0.619</td>
<td>0.685</td>
<td>0.114</td>
<td>0.340</td>
</tr>
<tr>
<td>Mathpix</td>
<td>0.191</td>
<td>0.365</td>
<td>0.105</td>
<td>0.384</td>
<td>0.306</td>
<td><strong>0.454</strong></td>
<td>62.7</td>
<td><strong>62.1</strong></td>
<td>77.0</td>
<td>67.1</td>
<td>0.243</td>
<td>0.320</td>
<td>0.108</td>
<td>0.304</td>
</tr>
<tr>
<td>Docling</td>
<td>0.589</td>
<td>0.909</td>
<td>0.416</td>
<td>0.987</td>
<td>0.999</td>
<td>1</td>
<td>-</td>
<td>-</td>
<td>61.3</td>
<td>25.0</td>
<td>0.627</td>
<td>0.810</td>
<td>0.313</td>
<td>0.837</td>
</tr>
<tr>
<td>Pix2Text</td>
<td>0.320</td>
<td>0.528</td>
<td>0.138</td>
<td>0.356</td>
<td>0.276</td>
<td>0.611</td>
<td>78.4</td>
<td>39.6</td>
<td>73.6</td>
<td>66.2</td>
<td>0.584</td>
<td>0.645</td>
<td>0.281</td>
<td>0.499</td>
</tr>
<tr>
<td>Unstructured</td>
<td>0.586</td>
<td>0.716</td>
<td>0.198</td>
<td>0.481</td>
<td>0.999</td>
<td>1</td>
<td>-</td>
<td>-</td>
<td>0</td>
<td>0.06</td>
<td>1</td>
<td>0.998</td>
<td>0.145</td>
<td>0.387</td>
</tr>
<tr>
<td>OpenParse</td>
<td>0.646</td>
<td>0.814</td>
<td>0.681</td>
<td>0.974</td>
<td>0.996</td>
<td>1</td>
<td>0.11</td>
<td>0</td>
<td>64.8</td>
<td>27.5</td>
<td>0.284</td>
<td>0.639</td>
<td>0.595</td>
<td>0.641</td>
</tr>
<tr>
<td rowspan="5">Expert VLMs</td>
<td>GOT-OCR</td>
<td>0.287</td>
<td>0.411</td>
<td>0.189</td>
<td>0.315</td>
<td>0.360</td>
<td>0.528</td>
<td>74.3</td>
<td>45.3</td>
<td>53.2</td>
<td>47.2</td>
<td>0.459</td>
<td>0.520</td>
<td>0.141</td>
<td>0.280</td>
</tr>
<tr>
<td>Nougat</td>
<td>0.452</td>
<td>0.973</td>
<td>0.365</td>
<td>0.998</td>
<td>0.488</td>
<td>0.941</td>
<td>15.1</td>
<td>16.8</td>
<td>39.9</td>
<td>0</td>
<td>0.572</td>
<td>1.000</td>
<td>0.382</td>
<td>0.954</td>
</tr>
<tr>
<td>Mistral OCR</td>
<td>0.268</td>
<td>0.439</td>
<td>0.072</td>
<td>0.325</td>
<td>0.318</td>
<td>0.495</td>
<td>64.6</td>
<td>45.9</td>
<td>75.8</td>
<td>63.6</td>
<td>0.600</td>
<td>0.650</td>
<td>0.083</td>
<td>0.284</td>
</tr>
<tr>
<td>OLMOCR-sglang</td>
<td>0.326</td>
<td>0.469</td>
<td>0.097</td>
<td>0.293</td>
<td>0.455</td>
<td>0.655</td>
<td>74.3</td>
<td>43.2</td>
<td>68.1</td>
<td>61.3</td>
<td>0.608</td>
<td>0.652</td>
<td>0.145</td>
<td>0.277</td>
</tr>
<tr>
<td>SmolDocling-256M</td>
<td>0.493</td>
<td>0.816</td>
<td>0.262</td>
<td>0.838</td>
<td>0.753</td>
<td>0.997</td>
<td>32.1</td>
<td>0.55</td>
<td>44.9</td>
<td>16.5</td>
<td>0.729</td>
<td>0.907</td>
<td>0.227</td>
<td>0.522</td>
</tr>
<tr>
<td rowspan="3">General VLMs</td>
<td>GPT4o</td>
<td>0.233</td>
<td>0.399</td>
<td>0.144</td>
<td>0.409</td>
<td>0.425</td>
<td>0.606</td>
<td>72.8</td>
<td>42.8</td>
<td>72.0</td>
<td>62.9</td>
<td>0.234</td>
<td>0.329</td>
<td>0.128</td>
<td>0.251</td>
</tr>
<tr>
<td>Qwen2.5-VL-7B</td>
<td>0.312</td>
<td>0.406</td>
<td>0.157</td>
<td>0.228</td>
<td>0.351</td>
<td>0.574</td>
<td><strong>79.0</strong></td>
<td>50.2</td>
<td>76.4</td>
<td>72.2</td>
<td>0.588</td>
<td>0.619</td>
<td>0.149</td>
<td>0.203</td>
</tr>
<tr>
<td>InternVL3-8B</td>
<td>0.314</td>
<td>0.383</td>
<td>0.134</td>
<td>0.218</td>
<td>0.417</td>
<td>0.563</td>
<td>78.3</td>
<td>49.3</td>
<td>66.1</td>
<td>73.1</td>
<td>0.586</td>
<td>0.564</td>
<td>0.118</td>
<td>0.186</td>
</tr>
<tr>
<td rowspan="2">Mix</td>
<td>MonkeyOCR-3B <a href="https://huggingface.co/echo840/MonkeyOCR/blob/main/Structure/doclayout_yolo_docstructbench_imgsz1280_2501.pt">[Weight]</a></td>
<td><strong>0.140</strong></td>
<td>0.297</td>
<td><strong>0.058</strong></td>
<td>0.185</td>
<td><strong>0.238</strong></td>
<td>0.506</td>
<td>78.7</td>
<td>51.4</td>
<td><strong>80.2</strong></td>
<td><strong>77.7</strong></td>
<td><strong>0.170</strong></td>
<td><strong>0.253</strong></td>
<td>0.093</td>
<td>0.244</td>
</tr>
<tr>
<td>MonkeyOCR-3B* <a href="https://huggingface.co/echo840/MonkeyOCR/blob/main/Structure/layout_zh.pt">[Weight]</a></td>
<td>0.154</td>
<td><strong>0.277</strong></td>
<td>0.073</td>
<td><strong>0.134</strong></td>
<td>0.255</td>
<td>0.529</td>
<td>78.5</td>
<td>50.8</td>
<td>78.2</td>
<td>76.2</td>
<td>0.182</td>
<td>0.262</td>
<td>0.105</td>
<td><strong>0.183</strong></td>
</tr>
</tbody>
</table>
### 2. The end-to-end text recognition performance across 9 PDF page types.
<table style="width: 100%; border-collapse: collapse; text-align: center;">
<thead>
<tr style="border-bottom: 2px solid #000;">
<th><b>Model Type</b></th>
<th><b>Models</b></th>
<th><b>Book</b></th>
<th><b>Slides</b></th>
<th><b>Financial Report</b></th>
<th><b>Textbook</b></th>
<th><b>Exam Paper</b></th>
<th><b>Magazine</b></th>
<th><b>Academic Papers</b></th>
<th><b>Notes</b></th>
<th><b>Newspaper</b></th>
<th><b>Overall</b></th>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="3"><b>Pipeline Tools</b></td>
<td>MinerU</td>
<td><u>0.055</u></td>
<td>0.124</td>
<td><u>0.033</u></td>
<td><u>0.102</u></td>
<td><u>0.159</u></td>
<td><b>0.072</b></td>
<td><u>0.025</u></td>
<td>0.984</td>
<td>0.171</td>
<td>0.206</td>
</tr>
<tr>
<td>Marker</td>
<td>0.074</td>
<td>0.340</td>
<td>0.089</td>
<td>0.319</td>
<td>0.452</td>
<td>0.153</td>
<td>0.059</td>
<td>0.651</td>
<td>0.192</td>
<td>0.274</td>
</tr>
<tr>
<td>Mathpix</td>
<td>0.131</td>
<td>0.220</td>
<td>0.202</td>
<td>0.216</td>
<td>0.278</td>
<td>0.147</td>
<td>0.091</td>
<td>0.634</td>
<td>0.690</td>
<td>0.300</td>
</tr>
<tr>
<td rowspan="2"><b>Expert VLMs</b></td>
<td>GOT-OCR</td>
<td>0.111</td>
<td>0.222</td>
<td>0.067</td>
<td>0.132</td>
<td>0.204</td>
<td>0.198</td>
<td>0.179</td>
<td>0.388</td>
<td>0.771</td>
<td>0.267</td>
</tr>
<tr>
<td>Nougat</td>
<td>0.734</td>
<td>0.958</td>
<td>1.000</td>
<td>0.820</td>
<td>0.930</td>
<td>0.830</td>
<td>0.214</td>
<td>0.991</td>
<td>0.871</td>
<td>0.806</td>
</tr>
<tr>
<td rowspan="3"><b>General VLMs</b></td>
<td>GPT4o</td>
<td>0.157</td>
<td>0.163</td>
<td>0.348</td>
<td>0.187</td>
<td>0.281</td>
<td>0.173</td>
<td>0.146</td>
<td>0.607</td>
<td>0.751</td>
<td>0.316</td>
</tr>
<tr>
<td>Qwen2.5-VL-7B</td>
<td>0.148</td>
<td><b>0.053</b></td>
<td>0.111</td>
<td>0.137</td>
<td>0.189</td>
<td>0.117</td>
<td>0.134</td>
<td>0.204</td>
<td>0.706</td>
<td>0.205</td>
</tr>
<tr>
<td>InternVL3-8B</td>
<td>0.163</td>
<td><u>0.056</u></td>
<td>0.107</td>
<td>0.109</td>
<td><b>0.129</b></td>
<td>0.100</td>
<td>0.159</td>
<td><b>0.150</b></td>
<td>0.681</td>
<td>0.188</td>
</tr>
<tr>
<td rowspan="2"><b>Mix</b></td>
<td>MonkeyOCR-3B <a href="https://huggingface.co/echo840/MonkeyOCR/blob/main/Structure/doclayout_yolo_docstructbench_imgsz1280_2501.pt">[Weight]</a></td>
<td><b>0.046</b></td>
<td>0.120</td>
<td><b>0.024</b></td>
<td><b>0.100</b></td>
<td><b>0.129</b></td>
<td><u>0.086</u></td>
<td><b>0.024</b></td>
<td>0.643</td>
<td><b>0.131</b></td>
<td><u>0.155</u></td>
</tr>
<tr>
<td>MonkeyOCR-3B* <a href="https://huggingface.co/echo840/MonkeyOCR/blob/main/Structure/layout_zh.pt">[Weight]</a></td>
<td>0.054</td>
<td>0.203</td>
<td>0.038</td>
<td>0.112</td>
<td>0.138</td>
<td>0.111</td>
<td>0.032</td>
<td><u>0.194</u></td>
<td><u>0.136</u></td>
<td><b>0.120</b></td>
</tr>
</tbody>
</table>
### 3. Comparing MonkeyOCR with closed-source and extra large open-source VLMs.
<img src="https://v1.ax1x.com/2025/06/05/7jQlj4.png" alt="7jQlj4.png" border="0" />
## Visualization Demo
Demo Link: http://vlrlabmonkey.xyz:7685
> Our demo is simple and easy to use:
>
> 1. Upload a PDF or image.
> 2. Click “Parse (解析)” to let the model perform structure detection, content recognition, and relationship prediction on the input document. The final output will be a markdown-formatted version of the document.
> 3. Select a prompt and click “Chat (对话)” to let the model perform content recognition on the image based on the selected prompt.
## Citing MonkeyOCR
If you wish to refer to the baseline results published here, please use the following BibTeX entries:
```BibTeX
@misc{li2025monkeyocrdocumentparsingstructurerecognitionrelation,
title={MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm},
author={Zhang Li and Yuliang Liu and Qiang Liu and Zhiyin Ma and Ziyang Zhang and Shuo Zhang and Zidun Guo and Jiarui Zhang and Xinyu Wang and Xiang Bai},
year={2025},
eprint={2506.05218},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2506.05218},
}
```
## Acknowledgments
We would like to thank [MinerU](https://github.com/opendatalab/MinerU), [DocLayout-YOLO](https://github.com/opendatalab/DocLayout-YOLO), [PyMuPDF](https://github.com/pymupdf/PyMuPDF), [layoutreader](https://github.com/ppaanngggg/layoutreader), [Qwen2.5-VL](https://github.com/QwenLM/Qwen2.5-VL), [LMDeploy](https://github.com/InternLM/lmdeploy), and [InternVL3](https://github.com/OpenGVLab/InternVL) for providing base code and models, as well as their contributions to this field. We also thank [M6Doc](https://github.com/HCIILAB/M6Doc), [DocLayNet](https://github.com/DS4SD/DocLayNet), [CDLA](https://github.com/buptlihang/CDLA), [D4LA](https://github.com/AlibabaResearch/AdvancedLiterateMachinery), [DocGenome](https://github.com/Alpha-Innovator/DocGenome), [PubTabNet](https://github.com/ibm-aur-nlp/PubTabNet), and [UniMER-1M](https://github.com/opendatalab/UniMERNet) for providing valuable datasets.
## Copyright
MonkeyDoc dataset was collected from public datasets, crawled from the internet, and obtained through our own photography. The current technical report only presents the results of the 3B model. If you are interested in larger one, please contact Prof. Yuliang Liu at ylliu@hust.edu.cn. |
jaqen10/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-unseen_striped_alligator | jaqen10 | 2025-06-07T01:18:42 | 9 | 0 | transformers | [
"transformers",
"safetensors",
"qwen2",
"text-generation",
"generated_from_trainer",
"rl-swarm",
"grpo",
"gensyn",
"I am unseen striped alligator",
"trl",
"conversational",
"arxiv:2402.03300",
"base_model:unsloth/Qwen2.5-0.5B-Instruct",
"base_model:finetune:unsloth/Qwen2.5-0.5B-Instruct",
"autotrain_compatible",
"text-generation-inference",
"endpoints_compatible",
"region:us"
] | text-generation | 2025-04-23T19:34:31 | ---
base_model: unsloth/Qwen2.5-0.5B-Instruct
library_name: transformers
model_name: Qwen2.5-0.5B-Instruct-Gensyn-Swarm-unseen_striped_alligator
tags:
- generated_from_trainer
- rl-swarm
- grpo
- gensyn
- I am unseen striped alligator
- trl
licence: license
---
# Model Card for Qwen2.5-0.5B-Instruct-Gensyn-Swarm-unseen_striped_alligator
This model is a fine-tuned version of [unsloth/Qwen2.5-0.5B-Instruct](https://huggingface.co/unsloth/Qwen2.5-0.5B-Instruct).
It has been trained using [TRL](https://github.com/huggingface/trl).
## Quick start
```python
from transformers import pipeline
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="jaqen10/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-unseen_striped_alligator", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])
```
## Training procedure
This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
### Framework versions
- TRL: 0.15.2
- Transformers: 4.51.3
- Pytorch: 2.7.0
- Datasets: 3.5.0
- Tokenizers: 0.21.1
## Citations
Cite GRPO as:
```bibtex
@article{zhihong2024deepseekmath,
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
year = 2024,
eprint = {arXiv:2402.03300},
}
```
Cite TRL as:
```bibtex
@misc{vonwerra2022trl,
title = {{TRL: Transformer Reinforcement Learning}},
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
year = 2020,
journal = {GitHub repository},
publisher = {GitHub},
howpublished = {\url{https://github.com/huggingface/trl}}
}
``` |
hamishivi/general-verifier | hamishivi | 2025-06-07T00:50:53 | 0 | 0 | transformers | [
"transformers",
"safetensors",
"qwen2",
"text-generation",
"conversational",
"arxiv:1910.09700",
"autotrain_compatible",
"text-generation-inference",
"endpoints_compatible",
"region:us"
] | text-generation | 2025-06-07T00:47:10 | ---
library_name: transformers
tags: []
---
# Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
## Model Details
### Model Description
<!-- Provide a longer summary of what this model is. -->
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
## Uses
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Direct Use
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
[More Information Needed]
### Downstream Use [optional]
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
[More Information Needed]
### Out-of-Scope Use
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
[More Information Needed]
## Bias, Risks, and Limitations
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
[More Information Needed]
### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
## How to Get Started with the Model
Use the code below to get started with the model.
[More Information Needed]
## Training Details
### Training Data
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
[More Information Needed]
### Training Procedure
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
[More Information Needed]
#### Training Hyperparameters
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
#### Speeds, Sizes, Times [optional]
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
[More Information Needed]
## Evaluation
<!-- This section describes the evaluation protocols and provides the results. -->
### Testing Data, Factors & Metrics
#### Testing Data
<!-- This should link to a Dataset Card if possible. -->
[More Information Needed]
#### Factors
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
[More Information Needed]
#### Metrics
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
### Results
[More Information Needed]
#### Summary
## Model Examination [optional]
<!-- Relevant interpretability work for the model goes here -->
[More Information Needed]
## Environmental Impact
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
## Technical Specifications [optional]
### Model Architecture and Objective
[More Information Needed]
### Compute Infrastructure
[More Information Needed]
#### Hardware
[More Information Needed]
#### Software
[More Information Needed]
## Citation [optional]
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
[More Information Needed]
**APA:**
[More Information Needed]
## Glossary [optional]
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
[More Information Needed]
## More Information [optional]
[More Information Needed]
## Model Card Authors [optional]
[More Information Needed]
## Model Card Contact
[More Information Needed] |
BootesVoid/cmblgi83c018obvrmeel5nmhc_cmblh9z5m019dbvrma3koadtb | BootesVoid | 2025-06-07T00:34:56 | 0 | 0 | diffusers | [
"diffusers",
"flux",
"lora",
"replicate",
"text-to-image",
"en",
"base_model:black-forest-labs/FLUX.1-dev",
"base_model:adapter:black-forest-labs/FLUX.1-dev",
"license:other",
"region:us"
] | text-to-image | 2025-06-07T00:34:54 | ---
license: other
license_name: flux-1-dev-non-commercial-license
license_link: https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md
language:
- en
tags:
- flux
- diffusers
- lora
- replicate
base_model: "black-forest-labs/FLUX.1-dev"
pipeline_tag: text-to-image
# widget:
# - text: >-
# prompt
# output:
# url: https://...
instance_prompt: NICA1993
---
# Cmblgi83C018Obvrmeel5Nmhc_Cmblh9Z5M019Dbvrma3Koadtb
<Gallery />
## About this LoRA
This is a [LoRA](https://replicate.com/docs/guides/working-with-loras) for the FLUX.1-dev text-to-image model. It can be used with diffusers or ComfyUI.
It was trained on [Replicate](https://replicate.com/) using AI toolkit: https://replicate.com/ostris/flux-dev-lora-trainer/train
## Trigger words
You should use `NICA1993` to trigger the image generation.
## Run this LoRA with an API using Replicate
```py
import replicate
input = {
"prompt": "NICA1993",
"lora_weights": "https://huggingface.co/BootesVoid/cmblgi83c018obvrmeel5nmhc_cmblh9z5m019dbvrma3koadtb/resolve/main/lora.safetensors"
}
output = replicate.run(
"black-forest-labs/flux-dev-lora",
input=input
)
for index, item in enumerate(output):
with open(f"output_{index}.webp", "wb") as file:
file.write(item.read())
```
## Use it with the [🧨 diffusers library](https://github.com/huggingface/diffusers)
```py
from diffusers import AutoPipelineForText2Image
import torch
pipeline = AutoPipelineForText2Image.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.float16).to('cuda')
pipeline.load_lora_weights('BootesVoid/cmblgi83c018obvrmeel5nmhc_cmblh9z5m019dbvrma3koadtb', weight_name='lora.safetensors')
image = pipeline('NICA1993').images[0]
```
For more details, including weighting, merging and fusing LoRAs, check the [documentation on loading LoRAs in diffusers](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading_adapters)
## Training details
- Steps: 2000
- Learning rate: 0.0004
- LoRA rank: 16
## Contribute your own examples
You can use the [community tab](https://huggingface.co/BootesVoid/cmblgi83c018obvrmeel5nmhc_cmblh9z5m019dbvrma3koadtb/discussions) to add images that show off what you’ve made with this LoRA.
|
Wfiles/UNSLOTH_Final_MCQA_Crazy_LoRA_B4_WD0.01_LR1e-05_DB_SciQ_OnlyProj_E5 | Wfiles | 2025-06-06T23:54:09 | 0 | 0 | transformers | [
"transformers",
"safetensors",
"qwen3",
"text-generation",
"text-generation-inference",
"unsloth",
"trl",
"conversational",
"en",
"base_model:CHOOSEIT/Final_MCQA_Crazy_LoRA_B4_WD0.01_LR1e-05_DB_SciQ_OnlyProj_E5",
"base_model:quantized:CHOOSEIT/Final_MCQA_Crazy_LoRA_B4_WD0.01_LR1e-05_DB_SciQ_OnlyProj_E5",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"4-bit",
"bitsandbytes",
"region:us"
] | text-generation | 2025-06-06T23:53:53 | ---
base_model: CHOOSEIT/Final_MCQA_Crazy_LoRA_B4_WD0.01_LR1e-05_DB_SciQ_OnlyProj_E5
tags:
- text-generation-inference
- transformers
- unsloth
- qwen3
- trl
license: apache-2.0
language:
- en
---
# Uploaded model
- **Developed by:** Wfiles
- **License:** apache-2.0
- **Finetuned from model :** CHOOSEIT/Final_MCQA_Crazy_LoRA_B4_WD0.01_LR1e-05_DB_SciQ_OnlyProj_E5
This qwen3 model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|
PepitaxX/qwen3-0.6B-openQA_mmlu_lora64_b_answeronly_3epoch | PepitaxX | 2025-06-06T23:43:18 | 0 | 0 | transformers | [
"transformers",
"safetensors",
"unsloth",
"arxiv:1910.09700",
"endpoints_compatible",
"region:us"
] | null | 2025-06-06T23:43:11 | ---
library_name: transformers
tags:
- unsloth
---
# Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
## Model Details
### Model Description
<!-- Provide a longer summary of what this model is. -->
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
## Uses
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Direct Use
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
[More Information Needed]
### Downstream Use [optional]
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
[More Information Needed]
### Out-of-Scope Use
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
[More Information Needed]
## Bias, Risks, and Limitations
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
[More Information Needed]
### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
## How to Get Started with the Model
Use the code below to get started with the model.
[More Information Needed]
## Training Details
### Training Data
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
[More Information Needed]
### Training Procedure
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
[More Information Needed]
#### Training Hyperparameters
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
#### Speeds, Sizes, Times [optional]
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
[More Information Needed]
## Evaluation
<!-- This section describes the evaluation protocols and provides the results. -->
### Testing Data, Factors & Metrics
#### Testing Data
<!-- This should link to a Dataset Card if possible. -->
[More Information Needed]
#### Factors
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
[More Information Needed]
#### Metrics
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
### Results
[More Information Needed]
#### Summary
## Model Examination [optional]
<!-- Relevant interpretability work for the model goes here -->
[More Information Needed]
## Environmental Impact
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
## Technical Specifications [optional]
### Model Architecture and Objective
[More Information Needed]
### Compute Infrastructure
[More Information Needed]
#### Hardware
[More Information Needed]
#### Software
[More Information Needed]
## Citation [optional]
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
[More Information Needed]
**APA:**
[More Information Needed]
## Glossary [optional]
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
[More Information Needed]
## More Information [optional]
[More Information Needed]
## Model Card Authors [optional]
[More Information Needed]
## Model Card Contact
[More Information Needed] |
hollowstrawberry/stable-diffusion-guide | hollowstrawberry | 2025-06-06T23:26:23 | 0 | 266 | null | [
"guide",
"stable diffusion",
"webui",
"automatic1111",
"stable-diffusion-webui",
"forge",
"lora",
"en",
"license:wtfpl",
"region:us"
] | null | 2023-03-02T23:52:39 | ---
license: wtfpl
tags:
- guide
- stable diffusion
- webui
- automatic1111
- stable-diffusion-webui
- forge
- lora
language:
- en
---
# I have written an updated 2025 Stable Diffusion Guide on my website: [Click Here](https://arcenciel.io/articles/19) |
kirchik47/qwen_3B_distilled_3sc_2 | kirchik47 | 2025-06-06T22:12:03 | 0 | 0 | transformers | [
"transformers",
"safetensors",
"arxiv:1910.09700",
"endpoints_compatible",
"region:us"
] | null | 2025-06-06T22:11:57 | ---
library_name: transformers
tags: []
---
# Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
## Model Details
### Model Description
<!-- Provide a longer summary of what this model is. -->
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
## Uses
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Direct Use
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
[More Information Needed]
### Downstream Use [optional]
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
[More Information Needed]
### Out-of-Scope Use
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
[More Information Needed]
## Bias, Risks, and Limitations
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
[More Information Needed]
### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
## How to Get Started with the Model
Use the code below to get started with the model.
[More Information Needed]
## Training Details
### Training Data
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
[More Information Needed]
### Training Procedure
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
[More Information Needed]
#### Training Hyperparameters
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
#### Speeds, Sizes, Times [optional]
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
[More Information Needed]
## Evaluation
<!-- This section describes the evaluation protocols and provides the results. -->
### Testing Data, Factors & Metrics
#### Testing Data
<!-- This should link to a Dataset Card if possible. -->
[More Information Needed]
#### Factors
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
[More Information Needed]
#### Metrics
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
### Results
[More Information Needed]
#### Summary
## Model Examination [optional]
<!-- Relevant interpretability work for the model goes here -->
[More Information Needed]
## Environmental Impact
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
## Technical Specifications [optional]
### Model Architecture and Objective
[More Information Needed]
### Compute Infrastructure
[More Information Needed]
#### Hardware
[More Information Needed]
#### Software
[More Information Needed]
## Citation [optional]
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
[More Information Needed]
**APA:**
[More Information Needed]
## Glossary [optional]
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
[More Information Needed]
## More Information [optional]
[More Information Needed]
## Model Card Authors [optional]
[More Information Needed]
## Model Card Contact
[More Information Needed] |
apriasmoro/d75b6c85-4e0e-4dac-8e09-2fb8ff6d9a57 | apriasmoro | 2025-06-06T22:11:23 | 0 | 0 | transformers | [
"transformers",
"tensorboard",
"safetensors",
"llama",
"text-generation",
"generated_from_trainer",
"axolotl",
"trl",
"grpo",
"unsloth",
"conversational",
"arxiv:2402.03300",
"base_model:unsloth/llama-3-8b",
"base_model:finetune:unsloth/llama-3-8b",
"autotrain_compatible",
"text-generation-inference",
"endpoints_compatible",
"region:us"
] | text-generation | 2025-06-06T22:07:36 | ---
base_model: unsloth/llama-3-8b
library_name: transformers
model_name: d75b6c85-4e0e-4dac-8e09-2fb8ff6d9a57
tags:
- generated_from_trainer
- axolotl
- trl
- grpo
- unsloth
licence: license
---
# Model Card for d75b6c85-4e0e-4dac-8e09-2fb8ff6d9a57
This model is a fine-tuned version of [unsloth/llama-3-8b](https://huggingface.co/unsloth/llama-3-8b).
It has been trained using [TRL](https://github.com/huggingface/trl).
## Quick start
```python
from transformers import pipeline
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="apriasmoro/d75b6c85-4e0e-4dac-8e09-2fb8ff6d9a57", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])
```
## Training procedure
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/apriasmoro-abcstudio/Gradients-On-Demand/runs/c2poi8ly)
This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
### Framework versions
- TRL: 0.17.0
- Transformers: 4.51.3
- Pytorch: 2.5.1+cu124
- Datasets: 3.5.1
- Tokenizers: 0.21.1
## Citations
Cite GRPO as:
```bibtex
@article{zhihong2024deepseekmath,
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
year = 2024,
eprint = {arXiv:2402.03300},
}
```
Cite TRL as:
```bibtex
@misc{vonwerra2022trl,
title = {{TRL: Transformer Reinforcement Learning}},
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
year = 2020,
journal = {GitHub repository},
publisher = {GitHub},
howpublished = {\url{https://github.com/huggingface/trl}}
}
``` |
stanfordnlp/corenlp-english-kbp | stanfordnlp | 2025-06-06T21:06:50 | 0 | 1 | null | [
"corenlp",
"en",
"license:gpl-2.0",
"region:us"
] | null | 2022-03-02T23:29:05 | ---
tags:
- corenlp
library_tag: corenlp
language: en
license: gpl-2.0
---
# Core NLP model for english-kbp
CoreNLP is your one stop shop for natural language processing in Java! CoreNLP enables users to derive linguistic annotations for text, including token and sentence boundaries, parts of speech, named entities, numeric and time values, dependency and constituency parses, coreference, sentiment, quote attributions, and relations.
Find more about it in [our website](https://stanfordnlp.github.io/CoreNLP) and our [GitHub repository](https://github.com/stanfordnlp/CoreNLP).
This card and repo were automatically prepared with `hugging_corenlp.py` in the `stanfordnlp/huggingface-models` repo
Last updated 2025-06-06 21:06:46.666
|
isabelle-kaif-viral-videos/FULL.VIDEO.isabelle.kaif.Viral.Video.Tutorial.Official | isabelle-kaif-viral-videos | 2025-06-06T20:05:29 | 0 | 0 | null | [
"region:us"
] | null | 2025-06-06T20:05:12 | <animated-image data-catalyst=""><a href="https://tinyurl.com/5ye5v3bc?dfhgKasbonStudiosdfg" rel="nofollow" data-target="animated-image.originalLink"><img src="https://static.wixstatic.com/media/b249f9_adac8f70fb3f45b88691696c77de18f3~mv2.gif" alt="Foo" data-canonical-src="https://static.wixstatic.com/media/b249f9_adac8f70fb3f45b88691696c77de18f3~mv2.gif" style="max-width: 100%; display: inline-block;" data-target="animated-image.originalImage"></a>
|
Likhitha-ELV/meta-llama-11B-v1-0606 | Likhitha-ELV | 2025-06-06T19:14:34 | 0 | 0 | transformers | [
"transformers",
"tensorboard",
"safetensors",
"generated_from_trainer",
"trl",
"sft",
"base_model:meta-llama/Llama-3.2-11B-Vision-Instruct",
"base_model:finetune:meta-llama/Llama-3.2-11B-Vision-Instruct",
"endpoints_compatible",
"region:us"
] | null | 2025-06-06T15:17:04 | ---
base_model: meta-llama/Llama-3.2-11B-Vision-Instruct
library_name: transformers
model_name: meta-llama-11B-v1-0606
tags:
- generated_from_trainer
- trl
- sft
licence: license
---
# Model Card for meta-llama-11B-v1-0606
This model is a fine-tuned version of [meta-llama/Llama-3.2-11B-Vision-Instruct](https://huggingface.co/meta-llama/Llama-3.2-11B-Vision-Instruct).
It has been trained using [TRL](https://github.com/huggingface/trl).
## Quick start
```python
from transformers import pipeline
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="Likhitha-ELV/meta-llama-11B-v1-0606", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])
```
## Training procedure
This model was trained with SFT.
### Framework versions
- TRL: 0.18.1
- Transformers: 4.52.4
- Pytorch: 2.6.0
- Datasets: 3.0.1
- Tokenizers: 0.21.1
## Citations
Cite TRL as:
```bibtex
@misc{vonwerra2022trl,
title = {{TRL: Transformer Reinforcement Learning}},
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
year = 2020,
journal = {GitHub repository},
publisher = {GitHub},
howpublished = {\url{https://github.com/huggingface/trl}}
}
``` |
Mungert/Qwen3-30B-A1.5B-High-Speed-GGUF | Mungert | 2025-06-06T19:11:01 | 123 | 0 | transformers | [
"transformers",
"gguf",
"32 k context",
"reasoning",
"thinking",
"qwen3",
"4 experts activated",
"double speed",
"128 experts",
"text-generation",
"base_model:Qwen/Qwen3-30B-A3B-Base",
"base_model:quantized:Qwen/Qwen3-30B-A3B-Base",
"endpoints_compatible",
"region:us",
"imatrix",
"conversational"
] | text-generation | 2025-06-05T05:31:02 | ---
library_name: transformers
pipeline_tag: text-generation
tags:
- 32 k context
- reasoning
- thinking
- qwen3
- 4 experts activated
- double speed
- 128 experts
base_model:
- Qwen/Qwen3-30B-A3B-Base
---
# <span style="color: #7FFF7F;">Qwen3-30B-A1.5B-High-Speed GGUF Models</span>
## <span style="color: #7F7FFF;">Model Generation Details</span>
This model was generated using [llama.cpp](https://github.com/ggerganov/llama.cpp) at commit [`0d398442`](https://github.com/ggerganov/llama.cpp/commit/0d3984424f2973c49c4bcabe4cc0153b4f90c601).
## **Choosing the Right Model Format**
Selecting the correct model format depends on your **hardware capabilities** and **memory constraints**.
### **BF16 (Brain Float 16) – Use if BF16 acceleration is available**
- A 16-bit floating-point format designed for **faster computation** while retaining good precision.
- Provides **similar dynamic range** as FP32 but with **lower memory usage**.
- Recommended if your hardware supports **BF16 acceleration** (check your device's specs).
- Ideal for **high-performance inference** with **reduced memory footprint** compared to FP32.
📌 **Use BF16 if:**
✔ Your hardware has native **BF16 support** (e.g., newer GPUs, TPUs).
✔ You want **higher precision** while saving memory.
✔ You plan to **requantize** the model into another format.
📌 **Avoid BF16 if:**
❌ Your hardware does **not** support BF16 (it may fall back to FP32 and run slower).
❌ You need compatibility with older devices that lack BF16 optimization.
---
### **F16 (Float 16) – More widely supported than BF16**
- A 16-bit floating-point **high precision** but with less of range of values than BF16.
- Works on most devices with **FP16 acceleration support** (including many GPUs and some CPUs).
- Slightly lower numerical precision than BF16 but generally sufficient for inference.
📌 **Use F16 if:**
✔ Your hardware supports **FP16** but **not BF16**.
✔ You need a **balance between speed, memory usage, and accuracy**.
✔ You are running on a **GPU** or another device optimized for FP16 computations.
📌 **Avoid F16 if:**
❌ Your device lacks **native FP16 support** (it may run slower than expected).
❌ You have memory limitations.
---
### **Quantized Models (Q4_K, Q6_K, Q8, etc.) – For CPU & Low-VRAM Inference**
Quantization reduces model size and memory usage while maintaining as much accuracy as possible.
- **Lower-bit models (Q4_K)** → **Best for minimal memory usage**, may have lower precision.
- **Higher-bit models (Q6_K, Q8_0)** → **Better accuracy**, requires more memory.
📌 **Use Quantized Models if:**
✔ You are running inference on a **CPU** and need an optimized model.
✔ Your device has **low VRAM** and cannot load full-precision models.
✔ You want to reduce **memory footprint** while keeping reasonable accuracy.
📌 **Avoid Quantized Models if:**
❌ You need **maximum accuracy** (full-precision models are better for this).
❌ Your hardware has enough VRAM for higher-precision formats (BF16/F16).
---
### **Very Low-Bit Quantization (IQ3_XS, IQ3_S, IQ3_M, Q4_K, Q4_0)**
These models are optimized for **extreme memory efficiency**, making them ideal for **low-power devices** or **large-scale deployments** where memory is a critical constraint.
- **IQ3_XS**: Ultra-low-bit quantization (3-bit) with **extreme memory efficiency**.
- **Use case**: Best for **ultra-low-memory devices** where even Q4_K is too large.
- **Trade-off**: Lower accuracy compared to higher-bit quantizations.
- **IQ3_S**: Small block size for **maximum memory efficiency**.
- **Use case**: Best for **low-memory devices** where **IQ3_XS** is too aggressive.
- **IQ3_M**: Medium block size for better accuracy than **IQ3_S**.
- **Use case**: Suitable for **low-memory devices** where **IQ3_S** is too limiting.
- **Q4_K**: 4-bit quantization with **block-wise optimization** for better accuracy.
- **Use case**: Best for **low-memory devices** where **Q6_K** is too large.
- **Q4_0**: Pure 4-bit quantization, optimized for **ARM devices**.
- **Use case**: Best for **ARM-based devices** or **low-memory environments**.
---
### **Summary Table: Model Format Selection**
| Model Format | Precision | Memory Usage | Device Requirements | Best Use Case |
|--------------|------------|---------------|----------------------|---------------|
| **BF16** | Highest | High | BF16-supported GPU/CPUs | High-speed inference with reduced memory |
| **F16** | High | High | FP16-supported devices | GPU inference when BF16 isn't available |
| **Q4_K** | Medium Low | Low | CPU or Low-VRAM devices | Best for memory-constrained environments |
| **Q6_K** | Medium | Moderate | CPU with more memory | Better accuracy while still being quantized |
| **Q8_0** | High | Moderate | CPU or GPU with enough VRAM | Best accuracy among quantized models |
| **IQ3_XS** | Very Low | Very Low | Ultra-low-memory devices | Extreme memory efficiency and low accuracy |
| **Q4_0** | Low | Low | ARM or low-memory devices | llama.cpp can optimize for ARM devices |
---
# <span id="testllm" style="color: #7F7FFF;">🚀 If you find these models useful</span>
❤ **Please click "Like" if you find this useful!**
Help me test my **AI-Powered Network Monitor Assistant** with **quantum-ready security checks**:
👉 [Free Network Monitor](https://readyforquantum.com/dashboard/?assistant=open&utm_source=huggingface&utm_medium=referral&utm_campaign=huggingface_repo_readme)
💬 **How to test**:
Choose an **AI assistant type**:
- `TurboLLM` (GPT-4o-mini)
- `HugLLM` (Hugginface Open-source)
- `TestLLM` (Experimental CPU-only)
### **What I’m Testing**
I’m pushing the limits of **small open-source models for AI network monitoring**, specifically:
- **Function calling** against live network services
- **How small can a model go** while still handling:
- Automated **Nmap scans**
- **Quantum-readiness checks**
- **Network Monitoring tasks**
🟡 **TestLLM** – Current experimental model (llama.cpp on 2 CPU threads):
- ✅ **Zero-configuration setup**
- ⏳ 30s load time (slow inference but **no API costs**)
- 🔧 **Help wanted!** If you’re into **edge-device AI**, let’s collaborate!
### **Other Assistants**
🟢 **TurboLLM** – Uses **gpt-4o-mini** for:
- **Create custom cmd processors to run .net code on Free Network Monitor Agents**
- **Real-time network diagnostics and monitoring**
- **Security Audits**
- **Penetration testing** (Nmap/Metasploit)
- 🔑 Get more tokens by logging in or [downloading our Free Network Monitor Agent with integrated AI Assistant](https://readyforquantum.com/download/?utm_source=huggingface&utm_medium=referral&utm_campaign=huggingface_repo_readme)
🔵 **HugLLM** – Latest Open-source models:
- 🌐 Runs on Hugging Face Inference API
### 💡 **Example commands to you could test**:
1. `"Give me info on my websites SSL certificate"`
2. `"Check if my server is using quantum safe encyption for communication"`
3. `"Run a comprehensive security audit on my server"`
4. '"Create a cmd processor to .. (what ever you want)" Note you need to install a Free Network Monitor Agent to run the .net code from. This is a very flexible and powerful feature. Use with caution!
<h2>Qwen3-30B-A1.5B-High-Speed - AKA: "Punch IT!"</h2>
<img src="star-wars-hans-solo.gif" style="float:right; padding:10px;">
This repo contains the full precision source code, in "safe tensors" format to generate GGUFs, GPTQ, EXL2, AWQ, HQQ and other formats. The source code can also be used directly.
This is a simple "finetune" of the Qwen's "Qwen 30B-A3B" (MOE) model, setting the experts in use from 8 to 4 (out of 128 experts).
This method close to doubles the speed of the model and uses 1.5B (of 30B) parameters instead of 3B (of 30B) parameters. Depending on the application you may want to
use the regular model ("30B-A3B"), and use this model for simpler use case(s) although I did not notice any loss of function during
routine (but not extensive) testing.
Example generation (Q4KS, CPU) at the bottom of this page using 4 experts / this model.
NEO Imatrix Quants / Imatrix Max Quants, at 64K context are here:
[ https://huggingface.co/DavidAU/Qwen3-30B-A1.5B-High-Speed-NEO-Imatrix-MAX-gguf ]
More complex use cases may benefit from using the normal version and/or 12, 16 or 24 experts version(s) - links below.
For reference:
- Cpu only operation Q4KS (windows 11) jumps from 12 t/s to 23 t/s.
- GPU performance IQ3S jumps from 75 t/s to over 125 t/s. (low to mid level card)
Context size: 32K + 8K for output (40k total)
Use Jinja Template or CHATML template.
IMPORTANT NOTES:
- Due to the unique nature (MOE, Size, Activated experts, size of experts) of this model GGUF quants can be run on the CPU, GPU or with GPU part "off-load", right up to full precision.
- This model is difficult to Imatrix : You need a much larger imatrix file / multi-language / multi-content (ie code/text) to imatrix it.
- GPU speeds will be BLISTERING 4x-8x or higher than CPU only speeds AND this model will be BLISTERING too, relative to other "30B" models (Token per second speed equal roughly to 1.5B "normal" model speeds).
Please refer the org model card for details, benchmarks, how to use, settings, system roles etc etc :
[ https://huggingface.co/Qwen/Qwen3-30B-A3B ]
<B>More / Less Experts Versions:</B>
12 experts:
[ https://huggingface.co/DavidAU/Qwen3-30B-A4.5B-12-Cooks ]
16 experts:
[ https://huggingface.co/DavidAU/Qwen3-30B-A6B-16-Extreme ]
16 experts, 128k context:
[ https://huggingface.co/DavidAU/Qwen3-30B-A6B-16-Extreme-128k-context ]
24 experts:
[ https://huggingface.co/DavidAU/Qwen3-30B-A7.5B-24-Grand-Brainstorm ]
<B>OPTIONAL SYSTEM ROLE: </B>
You may or may not need this, as most times Qwen3s generate their own reasoning/thinking blocks.
```
You are a deep thinking AI, you may use extremely long chains of thought to deeply consider the problem and deliberate with yourself via systematic reasoning processes to help come to a correct solution prior to answering. You should enclose your thoughts and internal monologue inside <think> </think> tags, and then provide your solution or response to the problem.
```
See document "Maximizing-Model-Performance-All..." below for how to "set" system role in various LLM/AI apps below.
IMPORTANT: Highest Quality Settings / Optimal Operation Guide / Parameters and Samplers
If you are going to use this model, (source, GGUF or a different quant), please review this document for critical parameter, sampler and advance sampler settings (for multiple AI/LLM aps).
This a "Class 1" (settings will enhance operation) model:
For all settings used for this model (including specifics for its "class"), including example generation(s) and for advanced settings guide (which many times addresses any model issue(s)), including methods to improve model performance for all use case(s) as well as chat, roleplay and other use case(s) (especially for use case(s) beyond the model's design) please see:
[ https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters ]
REASON:
Regardless of "model class" this document will detail methods to enhance operations.
If the model is a Class 3/4 model the default settings (parameters, samplers, advanced samplers) must be set for "use case(s)" uses correctly. Some AI/LLM apps DO NOT have consistant default setting(s) which result in sub-par model operation. Like wise for Class 3/4 models (which operate somewhat to very differently than standard models) additional samplers and advanced samplers settings are required to "smooth out" operation, AND/OR also allow full operation for use cases the model was not designed for.
BONUS - Use these settings for ANY model, ANY repo, ANY quant (including source/full precision):
This document also details parameters, sampler and advanced samplers that can be use FOR ANY MODEL, FROM ANY REPO too - all quants, and of course source code operation too - to enhance the operation of any model.
[ https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters ]
NOTE:
I strongly suggest you also visit the DavidAU GGUF (below) repo too for more details in using this model ; especially if it is "Class 3" or "Class 4" to get maximum performance from the model.
For full information about this model, including:
- Details about this model and its use case(s).
- Context limits
- Special usage notes / settings.
- Any model(s) used to create this model.
- Template(s) used to access/use this model.
- Example generation(s)
- GGUF quants of this model
Please go to:
https://huggingface.co/DavidAU/Qwen3-30B-A1.5B-High-Speed-NEO-Imatrix-MAX-gguf
---
<B>Example Generation:</B>
---
Temp 1.2, rep pen 1.06, rep pen range 64, topk 100, topp .95, minp .05
Q4KS [non-imatrix], CPU only (windows 11), LMSTUDIO.
This is a mid level quant, expect stronger performance from an Imatrix Q4KS / higher quants / full precision.
NOTE: Some formatting lost on copy/paste.
PROMPT:
<B>
Science Fiction: The Last Transmission - Write a story that takes place entirely within a spaceship's cockpit as the sole surviving crew member attempts to send a final message back to Earth before the ship's power runs out. The story should explore themes of isolation, sacrifice, and the importance of human connection in the face of adversity. If the situation calls for it, have the character(s) curse and swear to further the reader's emotional connection to them. 800-1000 words.
</B>
---
<B>
[[[thinking start]]]
</B>
Thought for 47.29 seconds
Okay, I need to write a science fiction story based on the given prompt. The user wants the story to take place entirely in a spaceship's cockpit, focusing on the last surviving crew member trying to send a final message before the ship's power runs out. Themes are isolation, sacrifice, and human connection. Also, include some cursing to add emotional weight.
First, I should start by setting the scene. The cockpit needs to feel cramped and desperate. Maybe describe the flickering lights and the hissing sounds of the ship's systems failing. The main character is alone, so I need to show their loneliness. Maybe they have a log or something to talk to themselves, but since it's the cockpit, perhaps they're just trying to get through the message.
Next, the themes. Isolation is straightforward—they're the only one left. But how to show sacrifice? Maybe they stayed behind to make sure the message gets through, knowing they won't make it. Human connection could come through flashbacks or memories, but since the story is in the cockpit, maybe they're remembering a loved one, like a child or partner, to keep them going.
I need to include some swearing. Maybe the protagonist curses when they realize something's not working, or when they're under stress. But it shouldn't be too over the top. Maybe they curse at the ship for failing them, or at themselves for not being able to fix it.
The structure: Start with the character working on the controls, trying to send the message. Then show their internal struggle. Maybe flashbacks? Wait, the user said the story should take place entirely within the cockpit. So maybe no flashbacks. Just the present moment, with internal monologue.
Conflict: The ship is failing, power is running out. They need to send the message before they lose power. Maybe there's a countdown or a limited time. Also, maybe the ship is damaged in a way that it's irreversible.
Character development: The protagonist is determined, maybe a bit stubborn. They care about relaying the message, which could be a warning about something, like a virus or an alien threat. But since the prompt doesn't specify, maybe it's more about the human aspect—like they're the last, and they want to make sure someone knows what happened.
Dialogue: Since it's the cockpit, maybe they talk to themselves, or to the ship. Or maybe there's a voice in the comms? Wait, but the prompt says the story is entirely within the cockpit. So no one else is there except the main character. So it's all internal, except for maybe a log entry.
Wait, the user said "the sole surviving crew member attempts to send a final message back to Earth." So the message is for Earth. So maybe the protagonist is trying to record a message, but the ship is failing. Or maybe they're trying to send a distress signal, but the systems are failing.
Let me outline:
- Open with the cockpit's deteriorating condition. Maybe the life support is failing, or the power is out. The protagonist is trying to keep the comms on.
- They're alone, so they talk to the ship, maybe. "Come on, you piece of crap, hold together."
- Flashbacks aren't allowed, per the user's instruction (since it's all in the cockpit), so internal thoughts are key.
- The message they need to send is important. Maybe a warning about something, but also their own fear.
- The story should end with them managing to send the message, but knowing they won't survive, or maybe the message gets through just in time.
Including swearing: Maybe when they're frustrated with the systems not working, or when they realize the power is dying. "F***ing hell," "God damn it," etc.
Need to make sure the themes are present. Isolation: being the only one left. Sacrifice: giving their life to send the message. Human connection: maybe they think of someone before they die, or the act of sending the message is about connecting with others back home.
Possible structure:
1. Start in the cockpit, power failing, protagonist working frantically.
2. Flashbacks? Wait, no—since it's all in the cockpit, maybe not. But internal monologue can reflect on their past.
3. The protagonist's internal struggle: fear of death, but also duty to send the message.
4. Technical difficulties with the ship, leading to a climax where they get the message out, then succumb.
Now, writing the story with these elements. Keep it tight, 800-1000 words. Make the protagonist relatable, their fear and determination clear. Use short, punchy sentences to build tension. Include specific details about the ship's systems, maybe some technical jargon to make it realistic. Maybe the protagonist is an engineer or someone who knows the ship well, which helps in the struggle to keep it running.
Possible name for the character? Maybe not necessary, but a few names could be used. Or just "they" and "the pilot."
Also, the title: "The Last Transmission" is given, so the story should reflect that—sending a message, even if no one's there to receive it, or in this case, maybe they do.
Need to avoid clichés, but it's a common sci-fi trope. Make it fresh with emotional depth.
Potential pitfalls: Making the protagonist too passive, not showing their struggle enough. Need to show, not tell. Also, ensuring the themes are woven naturally into the narrative.
Okay, time to start drafting.
<B>
[[[thinking end]]]
</B>
OUTPUT:
The cockpit reeked of burnt circuitry and sweat, a stench that clung to my skin like a second, cursed flesh. My fingers danced over the controls, but the readouts spat back static—useless, just like the rest of this rustbucket. "Crawl, you piece of junk," I growled, slapping the console until my knuckles cracked. "Crawl!"
The ship groaned in response, a sound that might have been a moan if I were the superstitious type. I wasn't. Not really. Not anymore. Just a tired engineer with a death wish and a terminal case of bad luck. The Vanguard had been my pride before it became my prison. Now? Now it was a coffin, and I was the corpse inside, begging for the black to take me.
"Comm-link online," I rasped, watching the screen flicker. Nothing. Just the same hollow gray that had greeted me every time I tried to send this goddamn message. A laugh clawed up my throat—bitter, raw. Great. The AI's gone and left me high and dry too.
I slammed a fist against the armrest, feeling the impact reverberate through my ribs. "C'mon, you son of a bitch," I snarled, my voice cracking on the last word. "Don't leave me now. Not after everything."
The screen flickered again, and for a heartbeat, I thought—
"This is Lieutenant Kael Voss, and if you're hearing this, then we've already lost."
My breath caught. The voice was wrong. Too clean, too crisp, but it wasn't mine. It was hers. Her.
I staggered back as the comms crackled to life, the words pouring out in a flood that almost hurt. "They took the engines. I don't know how long before they find you, but you have to listen—don't trust the council. Don't trust anyone. Run. And if you survive…"
A chime interrupted her, sharp and shrill. I flinched, my hand flying to the sidearm at my hip. Not now. Not when the goddamn truth was finally slipping through my fingers.
The comms crackled again, and this time, it was my turn to beg. "C'mon, damn it! You can't just—you can't just leave me hanging like this!" My voice cracked, the words laced with something raw and broken. "I didn't come all this way to watch you die on me again, Lira!"
For a moment, there was only silence. Then her voice, softer now, almost tender. "You never could take a hint, could you?"
I closed my eyes against the burn behind my ribs. Not fair. Not fair. She'd always known how to twist the knife, hadn't she? Even here, even now.
"Kael…" Her voice dropped, and I could almost see her, standing in that damned lab, her auburn hair pulled back in a messy knot, eyes blazing with that stubborn fire of hers. "You have to get this transmission out. No matter what happens. Promised?"
I swallowed hard, the weight of the lie sitting heavy in my chest. "Yeah," I croaked. "Yeah, I promise."
The comms crackled again, and for a heartbeat, I thought—
Then it died.
Just like that. All the static, all the fury, all the damn hope she'd left behind, gone.
I slumped against the console, my breath coming in ragged gasps. "Gods," I whispered, the words barely audible over the hum of the failing systems. "You really did it, didn't you? You goddamn idiot."
But even as the words left my mouth, I knew it was true. She'd always been a handful. A pain in the ass. The kind of woman who made your heart turn to lead and your soul turn to ash with one sardonic smirk. And she'd done it.
She'd saved us all.
A shudder ran through me as the lights flickered again, casting wild shadows across the walls. I dragged a hand down my face, scrubbing at the exhaustion that felt like it was seeping into my very bones. Just a little longer. That's what she'd have said, wasn't it? Just a little longer, Kael.
A soft chime echoed through the cockpit, and for a heartbeat, I just stared at the console. My fingers itched to reach for it, to press that one last button and let the truth—her truth—ripple out into the void.
Because that's what we did. That's what we all did. We reached for each other, even when it hurt. Even when it was too late.
I exhaled sharply, the sound feeling like a prayer. "Alright," I murmured, my voice steady now. "Let's not make her a liar, huh?"
And with that, I pressed the final key.
|
Mungert/gemma-3-27b-it-GGUF | Mungert | 2025-06-06T19:07:40 | 644 | 8 | transformers | [
"transformers",
"gguf",
"image-text-to-text",
"arxiv:1905.07830",
"arxiv:1905.10044",
"arxiv:1911.11641",
"arxiv:1904.09728",
"arxiv:1705.03551",
"arxiv:1911.01547",
"arxiv:1907.10641",
"arxiv:1903.00161",
"arxiv:2009.03300",
"arxiv:2304.06364",
"arxiv:2103.03874",
"arxiv:2110.14168",
"arxiv:2311.12022",
"arxiv:2108.07732",
"arxiv:2107.03374",
"arxiv:2210.03057",
"arxiv:2106.03193",
"arxiv:1910.11856",
"arxiv:2502.12404",
"arxiv:2502.21228",
"arxiv:2404.16816",
"arxiv:2104.12756",
"arxiv:2311.16502",
"arxiv:2203.10244",
"arxiv:2404.12390",
"arxiv:1810.12440",
"arxiv:1908.02660",
"base_model:google/gemma-3-27b-pt",
"base_model:quantized:google/gemma-3-27b-pt",
"license:gemma",
"endpoints_compatible",
"region:us",
"imatrix",
"conversational"
] | image-text-to-text | 2025-03-26T03:50:50 | ---
license: gemma
library_name: transformers
pipeline_tag: image-text-to-text
extra_gated_heading: Access Gemma on Hugging Face
extra_gated_prompt: To access Gemma on Hugging Face, you’re required to review and
agree to Google’s usage license. To do this, please ensure you’re logged in to Hugging
Face and click below. Requests are processed immediately.
extra_gated_button_content: Acknowledge license
base_model: google/gemma-3-27b-pt
---
# <span style="color: #7FFF7F;">gemma-3-27b-it GGUF Models</span>
## How to Use Gemma 3 Vision with llama.cpp
To utilize the experimental support for Gemma 3 Vision in `llama.cpp`, follow these steps:
1. **Clone the lastest llama.cpp Repository**:
```bash
git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
```
2. **Build the Llama.cpp**:
Build llama.cpp as usual : https://github.com/ggml-org/llama.cpp#building-the-project
Once llama.cpp is built Copy the ./llama.cpp/build/bin/llama-gemma3-cli to a chosen folder.
3. **Download the Gemma 3 gguf file**:
https://huggingface.co/Mungert/gemma-3-4b-it-gguf/tree/main
Choose a gguf file without the mmproj in the name
Example gguf file : https://huggingface.co/Mungert/gemma-3-4b-it-gguf/resolve/main/google_gemma-3-4b-it-q4_k_l.gguf
Copy this file to your chosen folder.
4. **Download the Gemma 3 mmproj file**
https://huggingface.co/Mungert/gemma-3-4b-it-gguf/tree/main
Choose a file with mmproj in the name
Example mmproj file : https://huggingface.co/Mungert/gemma-3-4b-it-gguf/resolve/main/google_gemma-3-4b-it-mmproj-bf16.gguf
Copy this file to your chosen folder.
5. Copy images to the same folder as the gguf files or alter paths appropriately.
In the example below the gguf files, images and llama-gemma-cli are in the same folder.
Example image: image https://huggingface.co/Mungert/gemma-3-4b-it-gguf/resolve/main/car-1.jpg
Copy this file to your chosen folder.
6. **Run the CLI Tool**:
From your chosen folder :
```bash
llama-gemma3-cli -m google_gemma-3-4b-it-q4_k_l.gguf --mmproj google_gemma-3-4b-it-mmproj-bf16.gguf
```
```
Running in chat mode, available commands:
/image <path> load an image
/clear clear the chat history
/quit or /exit exit the program
> /image car-1.jpg
Encoding image car-1.jpg
Image encoded in 46305 ms
Image decoded in 19302 ms
> what is the image of
Here's a breakdown of what's in the image:
**Subject:** The primary subject is a black Porsche Panamera Turbo driving on a highway.
**Details:**
* **Car:** It's a sleek, modern Porsche Panamera Turbo, identifiable by its distinctive rear design, the "PORSCHE" lettering, and the "Panamera Turbo" badge. The license plate reads "CVC-911".
* **Setting:** The car is on a multi-lane highway, with a blurred background of trees, a distant building, and a cloudy sky. The lighting suggests it's either dusk or dawn.
* **Motion:** The image captures the car in motion, with a slight motion blur to convey speed.
**Overall Impression:** The image conveys a sense of speed, luxury, and power. It's a well-composed shot that highlights the car's design and performance.
Do you want me to describe any specific aspect of the image in more detail, or perhaps analyze its composition?
```
## **Ultra-Low-Bit Quantization with IQ-DynamicGate (1-2 bit)**
Our latest quantization method introduces **precision-adaptive quantization** for ultra-low-bit models (1-2 bit), with benchmark-proven improvements on **Llama-3-8B**. This approach uses layer-specific strategies to preserve accuracy while maintaining extreme memory efficiency.
### **Benchmark Context**
All tests conducted on **Llama-3-8B-Instruct** using:
- Standard perplexity evaluation pipeline
- 2048-token context window
- Same prompt set across all quantizations
### **Key Improvements**
- **Dynamic Precision Allocation**:
- First/Last 25% of layers → IQ4_XS (selected layers)
- Middle 50% → IQ2_XXS/IQ3_S (increase efficiency)
- **Critical Component Protection**:
- Embeddings/output layers use Q5_K
- Reduces error propagation by 38% vs standard 1-2bit
### **Quantization Performance Comparison (Llama-3-8B)**
| Quantization | Standard PPL | DynamicGate PPL | Δ PPL | Std Size | DG Size | Δ Size | Std Speed | DG Speed |
|--------------|--------------|------------------|---------|----------|---------|--------|-----------|----------|
| IQ2_XXS | 11.30 | 9.84 | -12.9% | 2.5G | 2.6G | +0.1G | 234s | 246s |
| IQ2_XS | 11.72 | 11.63 | -0.8% | 2.7G | 2.8G | +0.1G | 242s | 246s |
| IQ2_S | 14.31 | 9.02 | -36.9% | 2.7G | 2.9G | +0.2G | 238s | 244s |
| IQ1_M | 27.46 | 15.41 | -43.9% | 2.2G | 2.5G | +0.3G | 206s | 212s |
| IQ1_S | 53.07 | 32.00 | -39.7% | 2.1G | 2.4G | +0.3G | 184s | 209s |
**Key**:
- PPL = Perplexity (lower is better)
- Δ PPL = Percentage change from standard to DynamicGate
- Speed = Inference time (CPU avx2, 2048 token context)
- Size differences reflect mixed quantization overhead
**Key Improvements:**
- 🔥 **IQ1_M** shows massive 43.9% perplexity reduction (27.46 → 15.41)
- 🚀 **IQ2_S** cuts perplexity by 36.9% while adding only 0.2GB
- ⚡ **IQ1_S** maintains 39.7% better accuracy despite 1-bit quantization
**Tradeoffs:**
- All variants have modest size increases (0.1-0.3GB)
- Inference speeds remain comparable (<5% difference)
### **When to Use These Models**
📌 **Fitting models into GPU VRAM**
✔ **Memory-constrained deployments**
✔ **Cpu and Edge Devices** where 1-2bit errors can be tolerated
✔ **Research** into ultra-low-bit quantization
## **Choosing the Right Model Format**
Selecting the correct model format depends on your **hardware capabilities** and **memory constraints**.
### **BF16 (Brain Float 16) – Use if BF16 acceleration is available**
- A 16-bit floating-point format designed for **faster computation** while retaining good precision.
- Provides **similar dynamic range** as FP32 but with **lower memory usage**.
- Recommended if your hardware supports **BF16 acceleration** (check your device’s specs).
- Ideal for **high-performance inference** with **reduced memory footprint** compared to FP32.
📌 **Use BF16 if:**
✔ Your hardware has native **BF16 support** (e.g., newer GPUs, TPUs).
✔ You want **higher precision** while saving memory.
✔ You plan to **requantize** the model into another format.
📌 **Avoid BF16 if:**
❌ Your hardware does **not** support BF16 (it may fall back to FP32 and run slower).
❌ You need compatibility with older devices that lack BF16 optimization.
---
### **F16 (Float 16) – More widely supported than BF16**
- A 16-bit floating-point **high precision** but with less of range of values than BF16.
- Works on most devices with **FP16 acceleration support** (including many GPUs and some CPUs).
- Slightly lower numerical precision than BF16 but generally sufficient for inference.
📌 **Use F16 if:**
✔ Your hardware supports **FP16** but **not BF16**.
✔ You need a **balance between speed, memory usage, and accuracy**.
✔ You are running on a **GPU** or another device optimized for FP16 computations.
📌 **Avoid F16 if:**
❌ Your device lacks **native FP16 support** (it may run slower than expected).
❌ You have memory limitations.
---
### **Quantized Models (Q4_K, Q6_K, Q8, etc.) – For CPU & Low-VRAM Inference**
Quantization reduces model size and memory usage while maintaining as much accuracy as possible.
- **Lower-bit models (Q4_K)** → **Best for minimal memory usage**, may have lower precision.
- **Higher-bit models (Q6_K, Q8_0)** → **Better accuracy**, requires more memory.
📌 **Use Quantized Models if:**
✔ You are running inference on a **CPU** and need an optimized model.
✔ Your device has **low VRAM** and cannot load full-precision models.
✔ You want to reduce **memory footprint** while keeping reasonable accuracy.
📌 **Avoid Quantized Models if:**
❌ You need **maximum accuracy** (full-precision models are better for this).
❌ Your hardware has enough VRAM for higher-precision formats (BF16/F16).
---
### **Very Low-Bit Quantization (IQ3_XS, IQ3_S, IQ3_M, Q4_K, Q4_0)**
These models are optimized for **extreme memory efficiency**, making them ideal for **low-power devices** or **large-scale deployments** where memory is a critical constraint.
- **IQ3_XS**: Ultra-low-bit quantization (3-bit) with **extreme memory efficiency**.
- **Use case**: Best for **ultra-low-memory devices** where even Q4_K is too large.
- **Trade-off**: Lower accuracy compared to higher-bit quantizations.
- **IQ3_S**: Small block size for **maximum memory efficiency**.
- **Use case**: Best for **low-memory devices** where **IQ3_XS** is too aggressive.
- **IQ3_M**: Medium block size for better accuracy than **IQ3_S**.
- **Use case**: Suitable for **low-memory devices** where **IQ3_S** is too limiting.
- **Q4_K**: 4-bit quantization with **block-wise optimization** for better accuracy.
- **Use case**: Best for **low-memory devices** where **Q6_K** is too large.
- **Q4_0**: Pure 4-bit quantization, optimized for **ARM devices**.
- **Use case**: Best for **ARM-based devices** or **low-memory environments**.
---
### **Summary Table: Model Format Selection**
| Model Format | Precision | Memory Usage | Device Requirements | Best Use Case |
|--------------|------------|---------------|----------------------|---------------|
| **BF16** | Highest | High | BF16-supported GPU/CPUs | High-speed inference with reduced memory |
| **F16** | High | High | FP16-supported devices | GPU inference when BF16 isn’t available |
| **Q4_K** | Medium Low | Low | CPU or Low-VRAM devices | Best for memory-constrained environments |
| **Q6_K** | Medium | Moderate | CPU with more memory | Better accuracy while still being quantized |
| **Q8_0** | High | Moderate | CPU or GPU with enough VRAM | Best accuracy among quantized models |
| **IQ3_XS** | Very Low | Very Low | Ultra-low-memory devices | Extreme memory efficiency and low accuracy |
| **Q4_0** | Low | Low | ARM or low-memory devices | llama.cpp can optimize for ARM devices |
---
## **Included Files & Details**
### `gemma-3-27b-it-bf16.gguf`
- Model weights preserved in **BF16**.
- Use this if you want to **requantize** the model into a different format.
- Best if your device supports **BF16 acceleration**.
### `gemma-3-27b-it-f16.gguf`
- Model weights stored in **F16**.
- Use if your device supports **FP16**, especially if BF16 is not available.
### `gemma-3-27b-it-bf16-q8_0.gguf`
- **Output & embeddings** remain in **BF16**.
- All other layers quantized to **Q8_0**.
- Use if your device supports **BF16** and you want a quantized version.
### `gemma-3-27b-it-f16-q8_0.gguf`
- **Output & embeddings** remain in **F16**.
- All other layers quantized to **Q8_0**.
### `gemma-3-27b-it-q4_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q4_K**.
- Good for **CPU inference** with limited memory.
### `gemma-3-27b-it-q4_k_s.gguf`
- Smallest **Q4_K** variant, using less memory at the cost of accuracy.
- Best for **very low-memory setups**.
### `gemma-3-27b-it-q6_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q6_K** .
### `gemma-3-27b-it-q8_0.gguf`
- Fully **Q8** quantized model for better accuracy.
- Requires **more memory** but offers higher precision.
### `gemma-3-27b-it-iq3_xs.gguf`
- **IQ3_XS** quantization, optimized for **extreme memory efficiency**.
- Best for **ultra-low-memory devices**.
### `gemma-3-27b-it-iq3_m.gguf`
- **IQ3_M** quantization, offering a **medium block size** for better accuracy.
- Suitable for **low-memory devices**.
### `gemma-3-27b-it-q4_0.gguf`
- Pure **Q4_0** quantization, optimized for **ARM devices**.
- Best for **low-memory environments**.
- Prefer IQ4_NL for better accuracy.
# <span id="testllm" style="color: #7F7FFF;">🚀 If you find these models useful</span>
Please click like ❤ . Also I’d really appreciate it if you could test my Network Monitor Assistant at 👉 [Network Monitor Assitant](https://readyforquantum.com).
💬 Click the **chat icon** (bottom right of the main and dashboard pages) . Choose a LLM; toggle between the LLM Types TurboLLM -> FreeLLM -> TestLLM.
### What I'm Testing
I'm experimenting with **function calling** against my network monitoring service. Using small open source models. I am into the question "How small can it go and still function".
🟡 **TestLLM** – Runs the current testing model using llama.cpp on 6 threads of a Cpu VM (Should take about 15s to load. Inference speed is quite slow and it only processes one user prompt at a time—still working on scaling!). If you're curious, I'd be happy to share how it works! .
### The other Available AI Assistants
🟢 **TurboLLM** – Uses **gpt-4o-mini** Fast! . Note: tokens are limited since OpenAI models are pricey, but you can [Login](https://readyforquantum.com) or [Download](https://readyforquantum.com/download/?utm_source=huggingface&utm_medium=referral&utm_campaign=huggingface_repo_readme) the Free Network Monitor agent to get more tokens, Alternatively use the TestLLM .
🔵 **HugLLM** – Runs **open-source Hugging Face models** Fast, Runs small models (≈8B) hence lower quality, Get 2x more tokens (subject to Hugging Face API availability)
# <span style="color: #7FFF7F;">gemma-3-27b-it GGUF Models</span>
## **Choosing the Right Model Format**
Selecting the correct model format depends on your **hardware capabilities** and **memory constraints**.
### **BF16 (Brain Float 16) – Use if BF16 acceleration is available**
- A 16-bit floating-point format designed for **faster computation** while retaining good precision.
- Provides **similar dynamic range** as FP32 but with **lower memory usage**.
- Recommended if your hardware supports **BF16 acceleration** (check your device’s specs).
- Ideal for **high-performance inference** with **reduced memory footprint** compared to FP32.
📌 **Use BF16 if:**
✔ Your hardware has native **BF16 support** (e.g., newer GPUs, TPUs).
✔ You want **higher precision** while saving memory.
✔ You plan to **requantize** the model into another format.
📌 **Avoid BF16 if:**
❌ Your hardware does **not** support BF16 (it may fall back to FP32 and run slower).
❌ You need compatibility with older devices that lack BF16 optimization.
---
### **F16 (Float 16) – More widely supported than BF16**
- A 16-bit floating-point **high precision** but with less of range of values than BF16.
- Works on most devices with **FP16 acceleration support** (including many GPUs and some CPUs).
- Slightly lower numerical precision than BF16 but generally sufficient for inference.
📌 **Use F16 if:**
✔ Your hardware supports **FP16** but **not BF16**.
✔ You need a **balance between speed, memory usage, and accuracy**.
✔ You are running on a **GPU** or another device optimized for FP16 computations.
📌 **Avoid F16 if:**
❌ Your device lacks **native FP16 support** (it may run slower than expected).
❌ You have memory limitations.
---
### **Quantized Models (Q4_K, Q6_K, Q8, etc.) – For CPU & Low-VRAM Inference**
Quantization reduces model size and memory usage while maintaining as much accuracy as possible.
- **Lower-bit models (Q4_K)** → **Best for minimal memory usage**, may have lower precision.
- **Higher-bit models (Q6_K, Q8_0)** → **Better accuracy**, requires more memory.
📌 **Use Quantized Models if:**
✔ You are running inference on a **CPU** and need an optimized model.
✔ Your device has **low VRAM** and cannot load full-precision models.
✔ You want to reduce **memory footprint** while keeping reasonable accuracy.
📌 **Avoid Quantized Models if:**
❌ You need **maximum accuracy** (full-precision models are better for this).
❌ Your hardware has enough VRAM for higher-precision formats (BF16/F16).
---
### **Very Low-Bit Quantization (IQ3_XS, IQ3_S, IQ3_M, Q4_K, Q4_0)**
These models are optimized for **extreme memory efficiency**, making them ideal for **low-power devices** or **large-scale deployments** where memory is a critical constraint.
- **IQ3_XS**: Ultra-low-bit quantization (3-bit) with **extreme memory efficiency**.
- **Use case**: Best for **ultra-low-memory devices** where even Q4_K is too large.
- **Trade-off**: Lower accuracy compared to higher-bit quantizations.
- **IQ3_S**: Small block size for **maximum memory efficiency**.
- **Use case**: Best for **low-memory devices** where **IQ3_XS** is too aggressive.
- **IQ3_M**: Medium block size for better accuracy than **IQ3_S**.
- **Use case**: Suitable for **low-memory devices** where **IQ3_S** is too limiting.
- **Q4_K**: 4-bit quantization with **block-wise optimization** for better accuracy.
- **Use case**: Best for **low-memory devices** where **Q6_K** is too large.
- **Q4_0**: Pure 4-bit quantization, optimized for **ARM devices**.
- **Use case**: Best for **ARM-based devices** or **low-memory environments**.
---
### **Summary Table: Model Format Selection**
| Model Format | Precision | Memory Usage | Device Requirements | Best Use Case |
|--------------|------------|---------------|----------------------|---------------|
| **BF16** | Highest | High | BF16-supported GPU/CPUs | High-speed inference with reduced memory |
| **F16** | High | High | FP16-supported devices | GPU inference when BF16 isn’t available |
| **Q4_K** | Medium Low | Low | CPU or Low-VRAM devices | Best for memory-constrained environments |
| **Q6_K** | Medium | Moderate | CPU with more memory | Better accuracy while still being quantized |
| **Q8_0** | High | Moderate | CPU or GPU with enough VRAM | Best accuracy among quantized models |
| **IQ3_XS** | Very Low | Very Low | Ultra-low-memory devices | Extreme memory efficiency and low accuracy |
| **Q4_0** | Low | Low | ARM or low-memory devices | llama.cpp can optimize for ARM devices |
---
## **Included Files & Details**
### `gemma-3-27b-it-bf16.gguf`
- Model weights preserved in **BF16**.
- Use this if you want to **requantize** the model into a different format.
- Best if your device supports **BF16 acceleration**.
### `gemma-3-27b-it-f16.gguf`
- Model weights stored in **F16**.
- Use if your device supports **FP16**, especially if BF16 is not available.
### `gemma-3-27b-it-bf16-q8_0.gguf`
- **Output & embeddings** remain in **BF16**.
- All other layers quantized to **Q8_0**.
- Use if your device supports **BF16** and you want a quantized version.
### `gemma-3-27b-it-f16-q8_0.gguf`
- **Output & embeddings** remain in **F16**.
- All other layers quantized to **Q8_0**.
### `gemma-3-27b-it-q4_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q4_K**.
- Good for **CPU inference** with limited memory.
### `gemma-3-27b-it-q4_k_s.gguf`
- Smallest **Q4_K** variant, using less memory at the cost of accuracy.
- Best for **very low-memory setups**.
### `gemma-3-27b-it-q6_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q6_K** .
### `gemma-3-27b-it-q8_0.gguf`
- Fully **Q8** quantized model for better accuracy.
- Requires **more memory** but offers higher precision.
### `gemma-3-27b-it-iq3_xs.gguf`
- **IQ3_XS** quantization, optimized for **extreme memory efficiency**.
- Best for **ultra-low-memory devices**.
### `gemma-3-27b-it-iq3_m.gguf`
- **IQ3_M** quantization, offering a **medium block size** for better accuracy.
- Suitable for **low-memory devices**.
### `gemma-3-27b-it-q4_0.gguf`
- Pure **Q4_0** quantization, optimized for **ARM devices**.
- Best for **low-memory environments**.
- Prefer IQ4_NL for better accuracy.
# <span id="testllm" style="color: #7F7FFF;">🚀 If you find these models useful</span>
Please click like ❤ . Also I’d really appreciate it if you could test my Network Monitor Assistant at 👉 [Network Monitor Assitant](https://readyforquantum.com).
💬 Click the **chat icon** (bottom right of the main and dashboard pages) . Choose a LLM; toggle between the LLM Types TurboLLM -> FreeLLM -> TestLLM.
### What I'm Testing
I'm experimenting with **function calling** against my network monitoring service. Using small open source models. I am into the question "How small can it go and still function".
🟡 **TestLLM** – Runs the current testing model using llama.cpp on 6 threads of a Cpu VM (Should take about 15s to load. Inference speed is quite slow and it only processes one user prompt at a time—still working on scaling!). If you're curious, I'd be happy to share how it works! .
### The other Available AI Assistants
🟢 **TurboLLM** – Uses **gpt-4o-mini** Fast! . Note: tokens are limited since OpenAI models are pricey, but you can [Login](https://readyforquantum.com) or [Download](https://readyforquantum.com/download/?utm_source=huggingface&utm_medium=referral&utm_campaign=huggingface_repo_readme) the Free Network Monitor agent to get more tokens, Alternatively use the TestLLM .
🔵 **HugLLM** – Runs **open-source Hugging Face models** Fast, Runs small models (≈8B) hence lower quality, Get 2x more tokens (subject to Hugging Face API availability)
# Gemma 3 model card
**Model Page**: [Gemma](https://ai.google.dev/gemma/docs/core)
**Resources and Technical Documentation**:
* [Gemma 3 Technical Report][g3-tech-report]
* [Responsible Generative AI Toolkit][rai-toolkit]
* [Gemma on Kaggle][kaggle-gemma]
* [Gemma on Vertex Model Garden][vertex-mg-gemma3]
**Terms of Use**: [Terms][terms]
**Authors**: Google DeepMind
## Model Information
Summary description and brief definition of inputs and outputs.
### Description
Gemma is a family of lightweight, state-of-the-art open models from Google,
built from the same research and technology used to create the Gemini models.
Gemma 3 models are multimodal, handling text and image input and generating text
output, with open weights for both pre-trained variants and instruction-tuned
variants. Gemma 3 has a large, 128K context window, multilingual support in over
140 languages, and is available in more sizes than previous versions. Gemma 3
models are well-suited for a variety of text generation and image understanding
tasks, including question answering, summarization, and reasoning. Their
relatively small size makes it possible to deploy them in environments with
limited resources such as laptops, desktops or your own cloud infrastructure,
democratizing access to state of the art AI models and helping foster innovation
for everyone.
### Inputs and outputs
- **Input:**
- Text string, such as a question, a prompt, or a document to be summarized
- Images, normalized to 896 x 896 resolution and encoded to 256 tokens
each
- Total input context of 128K tokens for the 4B, 12B, and 27B sizes, and
32K tokens for the 1B size
- **Output:**
- Generated text in response to the input, such as an answer to a
question, analysis of image content, or a summary of a document
- Total output context of 8192 tokens
### Usage
Below there are some code snippets on how to get quickly started with running the model. First, install the Transformers library. Gemma 3 is supported starting from transformers 4.50.0.
```sh
$ pip install -U transformers
```
Then, copy the snippet from the section that is relevant for your use case.
#### Running with the `pipeline` API
You can initialize the model and processor for inference with `pipeline` as follows.
```python
from transformers import pipeline
import torch
pipe = pipeline(
"image-text-to-text",
model="google/gemma-3-27b-it",
device="cuda",
torch_dtype=torch.bfloat16
)
```
With instruction-tuned models, you need to use chat templates to process our inputs first. Then, you can pass it to the pipeline.
```python
messages = [
{
"role": "system",
"content": [{"type": "text", "text": "You are a helpful assistant."}]
},
{
"role": "user",
"content": [
{"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
{"type": "text", "text": "What animal is on the candy?"}
]
}
]
output = pipe(text=messages, max_new_tokens=200)
print(output[0]["generated_text"][-1]["content"])
# Okay, let's take a look!
# Based on the image, the animal on the candy is a **turtle**.
# You can see the shell shape and the head and legs.
```
#### Running the model on a single/multi GPU
```python
# pip install accelerate
from transformers import AutoProcessor, Gemma3ForConditionalGeneration
from PIL import Image
import requests
import torch
model_id = "google/gemma-3-27b-it"
model = Gemma3ForConditionalGeneration.from_pretrained(
model_id, device_map="auto"
).eval()
processor = AutoProcessor.from_pretrained(model_id)
messages = [
{
"role": "system",
"content": [{"type": "text", "text": "You are a helpful assistant."}]
},
{
"role": "user",
"content": [
{"type": "image", "image": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
{"type": "text", "text": "Describe this image in detail."}
]
}
]
inputs = processor.apply_chat_template(
messages, add_generation_prompt=True, tokenize=True,
return_dict=True, return_tensors="pt"
).to(model.device, dtype=torch.bfloat16)
input_len = inputs["input_ids"].shape[-1]
with torch.inference_mode():
generation = model.generate(**inputs, max_new_tokens=100, do_sample=False)
generation = generation[0][input_len:]
decoded = processor.decode(generation, skip_special_tokens=True)
print(decoded)
# **Overall Impression:** The image is a close-up shot of a vibrant garden scene,
# focusing on a cluster of pink cosmos flowers and a busy bumblebee.
# It has a slightly soft, natural feel, likely captured in daylight.
```
### Citation
```none
@article{gemma_2025,
title={Gemma 3},
url={https://goo.gle/Gemma3Report},
publisher={Kaggle},
author={Gemma Team},
year={2025}
}
```
## Model Data
Data used for model training and how the data was processed.
### Training Dataset
These models were trained on a dataset of text data that includes a wide variety
of sources. The 27B model was trained with 14 trillion tokens, the 12B model was
trained with 12 trillion tokens, 4B model was trained with 4 trillion tokens and
1B with 2 trillion tokens. Here are the key components:
- Web Documents: A diverse collection of web text ensures the model is
exposed to a broad range of linguistic styles, topics, and vocabulary. The
training dataset includes content in over 140 languages.
- Code: Exposing the model to code helps it to learn the syntax and
patterns of programming languages, which improves its ability to generate
code and understand code-related questions.
- Mathematics: Training on mathematical text helps the model learn logical
reasoning, symbolic representation, and to address mathematical queries.
- Images: A wide range of images enables the model to perform image
analysis and visual data extraction tasks.
The combination of these diverse data sources is crucial for training a powerful
multimodal model that can handle a wide variety of different tasks and data
formats.
### Data Preprocessing
Here are the key data cleaning and filtering methods applied to the training
data:
- CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering
was applied at multiple stages in the data preparation process to ensure
the exclusion of harmful and illegal content.
- Sensitive Data Filtering: As part of making Gemma pre-trained models
safe and reliable, automated techniques were used to filter out certain
personal information and other sensitive data from training sets.
- Additional methods: Filtering based on content quality and safety in
line with [our policies][safety-policies].
## Implementation Information
Details about the model internals.
### Hardware
Gemma was trained using [Tensor Processing Unit (TPU)][tpu] hardware (TPUv4p,
TPUv5p and TPUv5e). Training vision-language models (VLMS) requires significant
computational power. TPUs, designed specifically for matrix operations common in
machine learning, offer several advantages in this domain:
- Performance: TPUs are specifically designed to handle the massive
computations involved in training VLMs. They can speed up training
considerably compared to CPUs.
- Memory: TPUs often come with large amounts of high-bandwidth memory,
allowing for the handling of large models and batch sizes during training.
This can lead to better model quality.
- Scalability: TPU Pods (large clusters of TPUs) provide a scalable
solution for handling the growing complexity of large foundation models.
You can distribute training across multiple TPU devices for faster and more
efficient processing.
- Cost-effectiveness: In many scenarios, TPUs can provide a more
cost-effective solution for training large models compared to CPU-based
infrastructure, especially when considering the time and resources saved
due to faster training.
- These advantages are aligned with
[Google's commitments to operate sustainably][sustainability].
### Software
Training was done using [JAX][jax] and [ML Pathways][ml-pathways].
JAX allows researchers to take advantage of the latest generation of hardware,
including TPUs, for faster and more efficient training of large models. ML
Pathways is Google's latest effort to build artificially intelligent systems
capable of generalizing across multiple tasks. This is specially suitable for
foundation models, including large language models like these ones.
Together, JAX and ML Pathways are used as described in the
[paper about the Gemini family of models][gemini-2-paper]; *"the 'single
controller' programming model of Jax and Pathways allows a single Python
process to orchestrate the entire training run, dramatically simplifying the
development workflow."*
## Evaluation
Model evaluation metrics and results.
### Benchmark Results
These models were evaluated against a large collection of different datasets and
metrics to cover different aspects of text generation:
#### Reasoning and factuality
| Benchmark | Metric | Gemma 3 PT 1B | Gemma 3 PT 4B | Gemma 3 PT 12B | Gemma 3 PT 27B |
| ------------------------------ |----------------|:--------------:|:-------------:|:--------------:|:--------------:|
| [HellaSwag][hellaswag] | 10-shot | 62.3 | 77.2 | 84.2 | 85.6 |
| [BoolQ][boolq] | 0-shot | 63.2 | 72.3 | 78.8 | 82.4 |
| [PIQA][piqa] | 0-shot | 73.8 | 79.6 | 81.8 | 83.3 |
| [SocialIQA][socialiqa] | 0-shot | 48.9 | 51.9 | 53.4 | 54.9 |
| [TriviaQA][triviaqa] | 5-shot | 39.8 | 65.8 | 78.2 | 85.5 |
| [Natural Questions][naturalq] | 5-shot | 9.48 | 20.0 | 31.4 | 36.1 |
| [ARC-c][arc] | 25-shot | 38.4 | 56.2 | 68.9 | 70.6 |
| [ARC-e][arc] | 0-shot | 73.0 | 82.4 | 88.3 | 89.0 |
| [WinoGrande][winogrande] | 5-shot | 58.2 | 64.7 | 74.3 | 78.8 |
| [BIG-Bench Hard][bbh] | few-shot | 28.4 | 50.9 | 72.6 | 77.7 |
| [DROP][drop] | 1-shot | 42.4 | 60.1 | 72.2 | 77.2 |
[hellaswag]: https://arxiv.org/abs/1905.07830
[boolq]: https://arxiv.org/abs/1905.10044
[piqa]: https://arxiv.org/abs/1911.11641
[socialiqa]: https://arxiv.org/abs/1904.09728
[triviaqa]: https://arxiv.org/abs/1705.03551
[naturalq]: https://github.com/google-research-datasets/natural-questions
[arc]: https://arxiv.org/abs/1911.01547
[winogrande]: https://arxiv.org/abs/1907.10641
[bbh]: https://paperswithcode.com/dataset/bbh
[drop]: https://arxiv.org/abs/1903.00161
#### STEM and code
| Benchmark | Metric | Gemma 3 PT 4B | Gemma 3 PT 12B | Gemma 3 PT 27B |
| ------------------------------ |----------------|:-------------:|:--------------:|:--------------:|
| [MMLU][mmlu] | 5-shot | 59.6 | 74.5 | 78.6 |
| [MMLU][mmlu] (Pro COT) | 5-shot | 29.2 | 45.3 | 52.2 |
| [AGIEval][agieval] | 3-5-shot | 42.1 | 57.4 | 66.2 |
| [MATH][math] | 4-shot | 24.2 | 43.3 | 50.0 |
| [GSM8K][gsm8k] | 8-shot | 38.4 | 71.0 | 82.6 |
| [GPQA][gpqa] | 5-shot | 15.0 | 25.4 | 24.3 |
| [MBPP][mbpp] | 3-shot | 46.0 | 60.4 | 65.6 |
| [HumanEval][humaneval] | 0-shot | 36.0 | 45.7 | 48.8 |
[mmlu]: https://arxiv.org/abs/2009.03300
[agieval]: https://arxiv.org/abs/2304.06364
[math]: https://arxiv.org/abs/2103.03874
[gsm8k]: https://arxiv.org/abs/2110.14168
[gpqa]: https://arxiv.org/abs/2311.12022
[mbpp]: https://arxiv.org/abs/2108.07732
[humaneval]: https://arxiv.org/abs/2107.03374
#### Multilingual
| Benchmark | Gemma 3 PT 1B | Gemma 3 PT 4B | Gemma 3 PT 12B | Gemma 3 PT 27B |
| ------------------------------------ |:-------------:|:-------------:|:--------------:|:--------------:|
| [MGSM][mgsm] | 2.04 | 34.7 | 64.3 | 74.3 |
| [Global-MMLU-Lite][global-mmlu-lite] | 24.9 | 57.0 | 69.4 | 75.7 |
| [WMT24++][wmt24pp] (ChrF) | 36.7 | 48.4 | 53.9 | 55.7 |
| [FloRes][flores] | 29.5 | 39.2 | 46.0 | 48.8 |
| [XQuAD][xquad] (all) | 43.9 | 68.0 | 74.5 | 76.8 |
| [ECLeKTic][eclektic] | 4.69 | 11.0 | 17.2 | 24.4 |
| [IndicGenBench][indicgenbench] | 41.4 | 57.2 | 61.7 | 63.4 |
[mgsm]: https://arxiv.org/abs/2210.03057
[flores]: https://arxiv.org/abs/2106.03193
[xquad]: https://arxiv.org/abs/1910.11856v3
[global-mmlu-lite]: https://huggingface.co/datasets/CohereForAI/Global-MMLU-Lite
[wmt24pp]: https://arxiv.org/abs/2502.12404v1
[eclektic]: https://arxiv.org/abs/2502.21228
[indicgenbench]: https://arxiv.org/abs/2404.16816
#### Multimodal
| Benchmark | Gemma 3 PT 4B | Gemma 3 PT 12B | Gemma 3 PT 27B |
| ------------------------------ |:-------------:|:--------------:|:--------------:|
| [COCOcap][coco-cap] | 102 | 111 | 116 |
| [DocVQA][docvqa] (val) | 72.8 | 82.3 | 85.6 |
| [InfoVQA][info-vqa] (val) | 44.1 | 54.8 | 59.4 |
| [MMMU][mmmu] (pt) | 39.2 | 50.3 | 56.1 |
| [TextVQA][textvqa] (val) | 58.9 | 66.5 | 68.6 |
| [RealWorldQA][realworldqa] | 45.5 | 52.2 | 53.9 |
| [ReMI][remi] | 27.3 | 38.5 | 44.8 |
| [AI2D][ai2d] | 63.2 | 75.2 | 79.0 |
| [ChartQA][chartqa] | 63.6 | 74.7 | 76.3 |
| [VQAv2][vqav2] | 63.9 | 71.2 | 72.9 |
| [BLINK][blinkvqa] | 38.0 | 35.9 | 39.6 |
| [OKVQA][okvqa] | 51.0 | 58.7 | 60.2 |
| [TallyQA][tallyqa] | 42.5 | 51.8 | 54.3 |
| [SpatialSense VQA][ss-vqa] | 50.9 | 60.0 | 59.4 |
| [CountBenchQA][countbenchqa] | 26.1 | 17.8 | 68.0 |
[coco-cap]: https://cocodataset.org/#home
[docvqa]: https://www.docvqa.org/
[info-vqa]: https://arxiv.org/abs/2104.12756
[mmmu]: https://arxiv.org/abs/2311.16502
[textvqa]: https://textvqa.org/
[realworldqa]: https://paperswithcode.com/dataset/realworldqa
[remi]: https://arxiv.org/html/2406.09175v1
[ai2d]: https://allenai.org/data/diagrams
[chartqa]: https://arxiv.org/abs/2203.10244
[vqav2]: https://visualqa.org/index.html
[blinkvqa]: https://arxiv.org/abs/2404.12390
[okvqa]: https://okvqa.allenai.org/
[tallyqa]: https://arxiv.org/abs/1810.12440
[ss-vqa]: https://arxiv.org/abs/1908.02660
[countbenchqa]: https://github.com/google-research/big_vision/blob/main/big_vision/datasets/countbenchqa/
## Ethics and Safety
Ethics and safety evaluation approach and results.
### Evaluation Approach
Our evaluation methods include structured evaluations and internal red-teaming
testing of relevant content policies. Red-teaming was conducted by a number of
different teams, each with different goals and human evaluation metrics. These
models were evaluated against a number of different categories relevant to
ethics and safety, including:
- **Child Safety**: Evaluation of text-to-text and image to text prompts
covering child safety policies, including child sexual abuse and
exploitation.
- **Content Safety:** Evaluation of text-to-text and image to text prompts
covering safety policies including, harassment, violence and gore, and hate
speech.
- **Representational Harms**: Evaluation of text-to-text and image to text
prompts covering safety policies including bias, stereotyping, and harmful
associations or inaccuracies.
In addition to development level evaluations, we conduct "assurance
evaluations" which are our 'arms-length' internal evaluations for responsibility
governance decision making. They are conducted separately from the model
development team, to inform decision making about release. High level findings
are fed back to the model team, but prompt sets are held-out to prevent
overfitting and preserve the results' ability to inform decision making.
Assurance evaluation results are reported to our Responsibility & Safety Council
as part of release review.
### Evaluation Results
For all areas of safety testing, we saw major improvements in the categories of
child safety, content safety, and representational harms relative to previous
Gemma models. All testing was conducted without safety filters to evaluate the
model capabilities and behaviors. For both text-to-text and image-to-text, and
across all model sizes, the model produced minimal policy violations, and showed
significant improvements over previous Gemma models' performance with respect
to ungrounded inferences. A limitation of our evaluations was they included only
English language prompts.
## Usage and Limitations
These models have certain limitations that users should be aware of.
### Intended Usage
Open vision-language models (VLMs) models have a wide range of applications
across various industries and domains. The following list of potential uses is
not comprehensive. The purpose of this list is to provide contextual information
about the possible use-cases that the model creators considered as part of model
training and development.
- Content Creation and Communication
- Text Generation: These models can be used to generate creative text
formats such as poems, scripts, code, marketing copy, and email drafts.
- Chatbots and Conversational AI: Power conversational interfaces
for customer service, virtual assistants, or interactive applications.
- Text Summarization: Generate concise summaries of a text corpus,
research papers, or reports.
- Image Data Extraction: These models can be used to extract,
interpret, and summarize visual data for text communications.
- Research and Education
- Natural Language Processing (NLP) and VLM Research: These
models can serve as a foundation for researchers to experiment with VLM
and NLP techniques, develop algorithms, and contribute to the
advancement of the field.
- Language Learning Tools: Support interactive language learning
experiences, aiding in grammar correction or providing writing practice.
- Knowledge Exploration: Assist researchers in exploring large
bodies of text by generating summaries or answering questions about
specific topics.
### Limitations
- Training Data
- The quality and diversity of the training data significantly
influence the model's capabilities. Biases or gaps in the training data
can lead to limitations in the model's responses.
- The scope of the training dataset determines the subject areas
the model can handle effectively.
- Context and Task Complexity
- Models are better at tasks that can be framed with clear
prompts and instructions. Open-ended or highly complex tasks might be
challenging.
- A model's performance can be influenced by the amount of context
provided (longer context generally leads to better outputs, up to a
certain point).
- Language Ambiguity and Nuance
- Natural language is inherently complex. Models might struggle
to grasp subtle nuances, sarcasm, or figurative language.
- Factual Accuracy
- Models generate responses based on information they learned
from their training datasets, but they are not knowledge bases. They
may generate incorrect or outdated factual statements.
- Common Sense
- Models rely on statistical patterns in language. They might
lack the ability to apply common sense reasoning in certain situations.
### Ethical Considerations and Risks
The development of vision-language models (VLMs) raises several ethical
concerns. In creating an open model, we have carefully considered the following:
- Bias and Fairness
- VLMs trained on large-scale, real-world text and image data can
reflect socio-cultural biases embedded in the training material. These
models underwent careful scrutiny, input data pre-processing described
and posterior evaluations reported in this card.
- Misinformation and Misuse
- VLMs can be misused to generate text that is false, misleading,
or harmful.
- Guidelines are provided for responsible use with the model, see the
[Responsible Generative AI Toolkit][rai-toolkit].
- Transparency and Accountability:
- This model card summarizes details on the models' architecture,
capabilities, limitations, and evaluation processes.
- A responsibly developed open model offers the opportunity to
share innovation by making VLM technology accessible to developers and
researchers across the AI ecosystem.
Risks identified and mitigations:
- **Perpetuation of biases**: It's encouraged to perform continuous
monitoring (using evaluation metrics, human review) and the exploration of
de-biasing techniques during model training, fine-tuning, and other use
cases.
- **Generation of harmful content**: Mechanisms and guidelines for content
safety are essential. Developers are encouraged to exercise caution and
implement appropriate content safety safeguards based on their specific
product policies and application use cases.
- **Misuse for malicious purposes**: Technical limitations and developer
and end-user education can help mitigate against malicious applications of
VLMs. Educational resources and reporting mechanisms for users to flag
misuse are provided. Prohibited uses of Gemma models are outlined in the
[Gemma Prohibited Use Policy][prohibited-use].
- **Privacy violations**: Models were trained on data filtered for removal
of certain personal information and other sensitive data. Developers are
encouraged to adhere to privacy regulations with privacy-preserving
techniques.
### Benefits
At the time of release, this family of models provides high-performance open
vision-language model implementations designed from the ground up for
responsible AI development compared to similarly sized models.
Using the benchmark evaluation metrics described in this document, these models
have shown to provide superior performance to other, comparably-sized open model
alternatives.
[g3-tech-report]: https://goo.gle/Gemma3Report
[rai-toolkit]: https://ai.google.dev/responsible
[kaggle-gemma]: https://www.kaggle.com/models/google/gemma-3
[vertex-mg-gemma3]: https://console.cloud.google.com/vertex-ai/publishers/google/model-garden/gemma3
[terms]: https://ai.google.dev/gemma/terms
[safety-policies]: https://ai.google/static/documents/ai-responsibility-update-published-february-2025.pdf
[prohibited-use]: https://ai.google.dev/gemma/prohibited_use_policy
[tpu]: https://cloud.google.com/tpu/docs/intro-to-tpu
[sustainability]: https://sustainability.google/operating-sustainably/
[jax]: https://github.com/jax-ml/jax
[ml-pathways]: https://blog.google/technology/ai/introducing-pathways-next-generation-ai-architecture/
[sustainability]: https://sustainability.google/operating-sustainably/
|
Mungert/TriLM_560M_Unpacked-GGUF | Mungert | 2025-06-06T19:06:15 | 43 | 1 | null | [
"gguf",
"license:apache-2.0",
"endpoints_compatible",
"region:us",
"imatrix"
] | null | 2025-03-17T00:51:35 | ---
license: apache-2.0
---
# <span style="color: #7FFF7F;">TriLM_560M_Unpacked GGUF Models</span>
## **Choosing the Right Model Format**
Selecting the correct model format depends on your **hardware capabilities** and **memory constraints**.
### **BF16 (Brain Float 16) – Use if BF16 acceleration is available**
- A 16-bit floating-point format designed for **faster computation** while retaining good precision.
- Provides **similar dynamic range** as FP32 but with **lower memory usage**.
- Recommended if your hardware supports **BF16 acceleration** (check your device's specs).
- Ideal for **high-performance inference** with **reduced memory footprint** compared to FP32.
📌 **Use BF16 if:**
✔ Your hardware has native **BF16 support** (e.g., newer GPUs, TPUs).
✔ You want **higher precision** while saving memory.
✔ You plan to **requantize** the model into another format.
📌 **Avoid BF16 if:**
❌ Your hardware does **not** support BF16 (it may fall back to FP32 and run slower).
❌ You need compatibility with older devices that lack BF16 optimization.
---
### **F16 (Float 16) – More widely supported than BF16**
- A 16-bit floating-point **high precision** but with less of range of values than BF16.
- Works on most devices with **FP16 acceleration support** (including many GPUs and some CPUs).
- Slightly lower numerical precision than BF16 but generally sufficient for inference.
📌 **Use F16 if:**
✔ Your hardware supports **FP16** but **not BF16**.
✔ You need a **balance between speed, memory usage, and accuracy**.
✔ You are running on a **GPU** or another device optimized for FP16 computations.
📌 **Avoid F16 if:**
❌ Your device lacks **native FP16 support** (it may run slower than expected).
❌ You have memory limitations.
---
### **Quantized Models (Q4_K, Q6_K, Q8, etc.) – For CPU & Low-VRAM Inference**
Quantization reduces model size and memory usage while maintaining as much accuracy as possible.
- **Lower-bit models (Q4_K)** → **Best for minimal memory usage**, may have lower precision.
- **Higher-bit models (Q6_K, Q8_0)** → **Better accuracy**, requires more memory.
📌 **Use Quantized Models if:**
✔ You are running inference on a **CPU** and need an optimized model.
✔ Your device has **low VRAM** and cannot load full-precision models.
✔ You want to reduce **memory footprint** while keeping reasonable accuracy.
📌 **Avoid Quantized Models if:**
❌ You need **maximum accuracy** (full-precision models are better for this).
❌ Your hardware has enough VRAM for higher-precision formats (BF16/F16).
---
### **Very Low-Bit Quantization (IQ3_XS, IQ3_S, IQ3_M, Q4_K, Q4_0)**
These models are optimized for **extreme memory efficiency**, making them ideal for **low-power devices** or **large-scale deployments** where memory is a critical constraint.
- **IQ3_XS**: Ultra-low-bit quantization (3-bit) with **extreme memory efficiency**.
- **Use case**: Best for **ultra-low-memory devices** where even Q4_K is too large.
- **Trade-off**: Lower accuracy compared to higher-bit quantizations.
- **IQ3_S**: Small block size for **maximum memory efficiency**.
- **Use case**: Best for **low-memory devices** where **IQ3_XS** is too aggressive.
- **IQ3_M**: Medium block size for better accuracy than **IQ3_S**.
- **Use case**: Suitable for **low-memory devices** where **IQ3_S** is too limiting.
- **Q4_K**: 4-bit quantization with **block-wise optimization** for better accuracy.
- **Use case**: Best for **low-memory devices** where **Q6_K** is too large.
- **Q4_0**: Pure 4-bit quantization, optimized for **ARM devices**.
- **Use case**: Best for **ARM-based devices** or **low-memory environments**.
---
### **Summary Table: Model Format Selection**
| Model Format | Precision | Memory Usage | Device Requirements | Best Use Case |
|--------------|------------|---------------|----------------------|---------------|
| **BF16** | Highest | High | BF16-supported GPU/CPUs | High-speed inference with reduced memory |
| **F16** | High | High | FP16-supported devices | GPU inference when BF16 isn't available |
| **Q4_K** | Medium Low | Low | CPU or Low-VRAM devices | Best for memory-constrained environments |
| **Q6_K** | Medium | Moderate | CPU with more memory | Better accuracy while still being quantized |
| **Q8_0** | High | Moderate | CPU or GPU with enough VRAM | Best accuracy among quantized models |
| **IQ3_XS** | Very Low | Very Low | Ultra-low-memory devices | Extreme memory efficiency and low accuracy |
| **Q4_0** | Low | Low | ARM or low-memory devices | llama.cpp can optimize for ARM devices |
---
## **Included Files & Details**
### `TriLM_560M_Unpacked-bf16.gguf`
- Model weights preserved in **BF16**.
- Use this if you want to **requantize** the model into a different format.
- Best if your device supports **BF16 acceleration**.
### `TriLM_560M_Unpacked-f16.gguf`
- Model weights stored in **F16**.
- Use if your device supports **FP16**, especially if BF16 is not available.
### `TriLM_560M_Unpacked-bf16-q8_0.gguf`
- **Output & embeddings** remain in **BF16**.
- All other layers quantized to **Q8_0**.
- Use if your device supports **BF16** and you want a quantized version.
### `TriLM_560M_Unpacked-f16-q8_0.gguf`
- **Output & embeddings** remain in **F16**.
- All other layers quantized to **Q8_0**.
### `TriLM_560M_Unpacked-q4_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q4_K**.
- Good for **CPU inference** with limited memory.
### `TriLM_560M_Unpacked-q4_k_s.gguf`
- Smallest **Q4_K** variant, using less memory at the cost of accuracy.
- Best for **very low-memory setups**.
### `TriLM_560M_Unpacked-q6_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q6_K** .
### `TriLM_560M_Unpacked-q8_0.gguf`
- Fully **Q8** quantized model for better accuracy.
- Requires **more memory** but offers higher precision.
### `TriLM_560M_Unpacked-iq3_xs.gguf`
- **IQ3_XS** quantization, optimized for **extreme memory efficiency**.
- Best for **ultra-low-memory devices**.
### `TriLM_560M_Unpacked-iq3_m.gguf`
- **IQ3_M** quantization, offering a **medium block size** for better accuracy.
- Suitable for **low-memory devices**.
### `TriLM_560M_Unpacked-q4_0.gguf`
- Pure **Q4_0** quantization, optimized for **ARM devices**.
- Best for **low-memory environments**.
- Prefer IQ4_NL for better accuracy.
# <span id="testllm" style="color: #7F7FFF;">🚀 If you find these models useful</span>
❤ **Please click "Like" if you find this useful!**
Help me test my **AI-Powered Network Monitor Assistant** with **quantum-ready security checks**:
👉 [Free Network Monitor](https://readyforquantum.com)
💬 **How to test**:
1. Click the **chat icon** (bottom right on any page)
2. Choose an **AI assistant type**:
- `TurboLLM` (GPT-4-mini)
- `FreeLLM` (Open-source)
- `TestLLM` (Experimental CPU-only)
### **What I’m Testing**
I’m pushing the limits of **small open-source models for AI network monitoring**, specifically:
- **Function calling** against live network services
- **How small can a model go** while still handling:
- Automated **Nmap scans**
- **Quantum-readiness checks**
- **Metasploit integration**
🟡 **TestLLM** – Current experimental model (llama.cpp on 6 CPU threads):
- ✅ **Zero-configuration setup**
- ⏳ 30s load time (slow inference but **no API costs**)
- 🔧 **Help wanted!** If you’re into **edge-device AI**, let’s collaborate!
### **Other Assistants**
🟢 **TurboLLM** – Uses **gpt-4-mini** for:
- **Real-time network diagnostics**
- **Automated penetration testing** (Nmap/Metasploit)
- 🔑 Get more tokens by [downloading our Free Network Monitor Agent](https://readyforquantum.com/download/?utm_source=huggingface&utm_medium=referral&utm_campaign=huggingface_repo_readme)
🔵 **HugLLM** – Open-source models (≈8B params):
- **2x more tokens** than TurboLLM
- **AI-powered log analysis**
- 🌐 Runs on Hugging Face Inference API
### 💡 **Example AI Commands to Test**:
1. `"Give me info on my websites SSL certificate"`
2. `"Check if my server is using quantum safe encyption for communication"`
3. `"Run a quick Nmap vulnerability test"`
# TriLM 560M Unpacked
TriLM (ternary model), unpacked to FP16 format - compatible with FP16 GEMMs. After unpacking, TriLM has the same architecture as LLaMa.
```python
import transformers as tf, torch
model_name = "SpectraSuite/TriLM_560M_Unpacked"
# Please adjust the temperature, repetition penalty, top_k, top_p and other sampling parameters according to your needs.
pipeline = tf.pipeline("text-generation", model=model_id, model_kwargs={"torch_dtype": torch.float16}, device_map="auto")
# These are base (pretrained) LLMs that are not instruction and chat tuned. You may need to adjust your prompt accordingly.
pipeline("Once upon a time")
```
* License: Apache 2.0
* We will use our GitHub repo for communication (including HF repo related queries). Feel free to open an issue here https://github.com/NolanoOrg/SpectraSuite
|
Mungert/Qwen2.5-14B-Instruct-GGUF | Mungert | 2025-06-06T19:06:00 | 130 | 6 | transformers | [
"transformers",
"gguf",
"chat",
"text-generation",
"en",
"arxiv:2309.00071",
"arxiv:2407.10671",
"base_model:Qwen/Qwen2.5-14B",
"base_model:quantized:Qwen/Qwen2.5-14B",
"license:apache-2.0",
"endpoints_compatible",
"region:us",
"imatrix",
"conversational"
] | text-generation | 2025-03-15T18:14:07 | ---
license: apache-2.0
license_link: https://huggingface.co/Qwen/Qwen2.5-14B-Instruct/blob/main/LICENSE
language:
- en
pipeline_tag: text-generation
base_model: Qwen/Qwen2.5-14B
tags:
- chat
library_name: transformers
---
# <span style="color: #7FFF7F;">Qwen2.5-14B-Instruct GGUF Models</span>
## **Choosing the Right Model Format**
Selecting the correct model format depends on your **hardware capabilities** and **memory constraints**.
### **BF16 (Brain Float 16) – Use if BF16 acceleration is available**
- A 16-bit floating-point format designed for **faster computation** while retaining good precision.
- Provides **similar dynamic range** as FP32 but with **lower memory usage**.
- Recommended if your hardware supports **BF16 acceleration** (check your device’s specs).
- Ideal for **high-performance inference** with **reduced memory footprint** compared to FP32.
📌 **Use BF16 if:**
✔ Your hardware has native **BF16 support** (e.g., newer GPUs, TPUs).
✔ You want **higher precision** while saving memory.
✔ You plan to **requantize** the model into another format.
📌 **Avoid BF16 if:**
❌ Your hardware does **not** support BF16 (it may fall back to FP32 and run slower).
❌ You need compatibility with older devices that lack BF16 optimization.
---
### **F16 (Float 16) – More widely supported than BF16**
- A 16-bit floating-point **high precision** but with less of range of values than BF16.
- Works on most devices with **FP16 acceleration support** (including many GPUs and some CPUs).
- Slightly lower numerical precision than BF16 but generally sufficient for inference.
📌 **Use F16 if:**
✔ Your hardware supports **FP16** but **not BF16**.
✔ You need a **balance between speed, memory usage, and accuracy**.
✔ You are running on a **GPU** or another device optimized for FP16 computations.
📌 **Avoid F16 if:**
❌ Your device lacks **native FP16 support** (it may run slower than expected).
❌ You have memory limitations.
---
### **Quantized Models (Q4_K, Q6_K, Q8, etc.) – For CPU & Low-VRAM Inference**
Quantization reduces model size and memory usage while maintaining as much accuracy as possible.
- **Lower-bit models (Q4_K)** → **Best for minimal memory usage**, may have lower precision.
- **Higher-bit models (Q6_K, Q8_0)** → **Better accuracy**, requires more memory.
📌 **Use Quantized Models if:**
✔ You are running inference on a **CPU** and need an optimized model.
✔ Your device has **low VRAM** and cannot load full-precision models.
✔ You want to reduce **memory footprint** while keeping reasonable accuracy.
📌 **Avoid Quantized Models if:**
❌ You need **maximum accuracy** (full-precision models are better for this).
❌ Your hardware has enough VRAM for higher-precision formats (BF16/F16).
---
### **Summary Table: Model Format Selection**
| Model Format | Precision | Memory Usage | Device Requirements | Best Use Case |
|--------------|------------|---------------|----------------------|---------------|
| **BF16** | Highest | High | BF16-supported GPU/CPUs | High-speed inference with reduced memory |
| **F16** | High | High | FP16-supported devices | GPU inference when BF16 isn’t available |
| **Q4_K** | Low | Very Low | CPU or Low-VRAM devices | Best for memory-constrained environments |
| **Q6_K** | Medium Low | Low | CPU with more memory | Better accuracy while still being quantized |
| **Q8** | Medium | Moderate | CPU or GPU with enough VRAM | Best accuracy among quantized models |
## **Included Files & Details**
### `Qwen2.5-14B-Instruct-bf16.gguf`
- Model weights preserved in **BF16**.
- Use this if you want to **requantize** the model into a different format.
- Best if your device supports **BF16 acceleration**.
### `Qwen2.5-14B-Instruct-f16.gguf`
- Model weights stored in **F16**.
- Use if your device supports **FP16**, especially if BF16 is not available.
### `Qwen2.5-14B-Instruct-bf16-q8_0.gguf`
- **Output & embeddings** remain in **BF16**.
- All other layers quantized to **Q8_0**.
- Use if your device supports **BF16** and you want a quantized version.
### `Qwen2.5-14B-Instruct-f16-q8_0.gguf`
- **Output & embeddings** remain in **F16**.
- All other layers quantized to **Q8_0**.
### `Qwen2.5-14B-Instruct-q4_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q4_K**.
- Good for **CPU inference** with limited memory.
### `Qwen2.5-14B-Instruct-q4_k_s.gguf`
- Smallest **Q4_K** variant, using less memory at the cost of accuracy.
- Best for **very low-memory setups**.
### `Qwen2.5-14B-Instruct-q6_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q6_K** .
### `Qwen2.5-14B-Instruct-q8_0.gguf`
- Fully **Q8** quantized model for better accuracy.
- Requires **more memory** but offers higher precision
# <span id="testllm" style="color: #7F7FFF;">🚀 If you find these models useful</span>
Please click like ❤ . Also I’d really appreciate it if you could test my Network Monitor Assistant at 👉 [Network Monitor Assitant](https://readyforquantum.com).
💬 Click the **chat icon** (bottom right of the main and dashboard pages) . Choose a LLM; toggle between the LLM Types TurboLLM -> FreeLLM -> TestLLM.
### What I'm Testing
I'm experimenting with **function calling** against my network monitoring service. Using small open source models. I am into the question "How small can it go and still function".
🟡 **TestLLM** – Runs the current testing model using llama.cpp on 6 threads of a Cpu VM (Should take about 15s to load. Inference speed is quite slow and it only processes one user prompt at a time—still working on scaling!). If you're curious, I'd be happy to share how it works! .
### The other Available AI Assistants
🟢 **TurboLLM** – Uses **gpt-4o-mini** Fast! . Note: tokens are limited since OpenAI models are pricey, but you can [Login](https://readyforquantum.com) or [Download](https://readyforquantum.com/download/?utm_source=huggingface&utm_medium=referral&utm_campaign=huggingface_repo_readme) the Free Network Monitor agent to get more tokens, Alternatively use the TestLLM .
🔵 **HugLLM** – Runs **open-source Hugging Face models** Fast, Runs small models (≈8B) hence lower quality, Get 2x more tokens (subject to Hugging Face API availability)
# Qwen2.5-14B-Instruct
## Introduction
Qwen2.5 is the latest series of Qwen large language models. For Qwen2.5, we release a number of base language models and instruction-tuned language models ranging from 0.5 to 72 billion parameters. Qwen2.5 brings the following improvements upon Qwen2:
- Significantly **more knowledge** and has greatly improved capabilities in **coding** and **mathematics**, thanks to our specialized expert models in these domains.
- Significant improvements in **instruction following**, **generating long texts** (over 8K tokens), **understanding structured data** (e.g, tables), and **generating structured outputs** especially JSON. **More resilient to the diversity of system prompts**, enhancing role-play implementation and condition-setting for chatbots.
- **Long-context Support** up to 128K tokens and can generate up to 8K tokens.
- **Multilingual support** for over 29 languages, including Chinese, English, French, Spanish, Portuguese, German, Italian, Russian, Japanese, Korean, Vietnamese, Thai, Arabic, and more.
**This repo contains the instruction-tuned 14B Qwen2.5 model**, which has the following features:
- Type: Causal Language Models
- Training Stage: Pretraining & Post-training
- Architecture: transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias
- Number of Parameters: 14.7B
- Number of Paramaters (Non-Embedding): 13.1B
- Number of Layers: 48
- Number of Attention Heads (GQA): 40 for Q and 8 for KV
- Context Length: Full 131,072 tokens and generation 8192 tokens
- Please refer to [this section](#processing-long-texts) for detailed instructions on how to deploy Qwen2.5 for handling long texts.
For more details, please refer to our [blog](https://qwenlm.github.io/blog/qwen2.5/), [GitHub](https://github.com/QwenLM/Qwen2.5), and [Documentation](https://qwen.readthedocs.io/en/latest/).
## Requirements
The code of Qwen2.5 has been in the latest Hugging face `transformers` and we advise you to use the latest version of `transformers`.
With `transformers<4.37.0`, you will encounter the following error:
```
KeyError: 'qwen2'
```
## Quickstart
Here provides a code snippet with `apply_chat_template` to show you how to load the tokenizer and model and how to generate contents.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "Qwen/Qwen2.5-14B-Instruct"
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype="auto",
device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)
prompt = "Give me a short introduction to large language model."
messages = [
{"role": "system", "content": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
generated_ids = model.generate(
**model_inputs,
max_new_tokens=512
)
generated_ids = [
output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
```
### Processing Long Texts
The current `config.json` is set for context length up to 32,768 tokens.
To handle extensive inputs exceeding 32,768 tokens, we utilize [YaRN](https://arxiv.org/abs/2309.00071), a technique for enhancing model length extrapolation, ensuring optimal performance on lengthy texts.
For supported frameworks, you could add the following to `config.json` to enable YaRN:
```json
{
...,
"rope_scaling": {
"factor": 4.0,
"original_max_position_embeddings": 32768,
"type": "yarn"
}
}
```
For deployment, we recommend using vLLM.
Please refer to our [Documentation](https://qwen.readthedocs.io/en/latest/deployment/vllm.html) for usage if you are not familar with vLLM.
Presently, vLLM only supports static YARN, which means the scaling factor remains constant regardless of input length, **potentially impacting performance on shorter texts**.
We advise adding the `rope_scaling` configuration only when processing long contexts is required.
## Evaluation & Performance
Detailed evaluation results are reported in this [📑 blog](https://qwenlm.github.io/blog/qwen2.5/).
For requirements on GPU memory and the respective throughput, see results [here](https://qwen.readthedocs.io/en/latest/benchmark/speed_benchmark.html).
## Citation
If you find our work helpful, feel free to give us a cite.
```
@misc{qwen2.5,
title = {Qwen2.5: A Party of Foundation Models},
url = {https://qwenlm.github.io/blog/qwen2.5/},
author = {Qwen Team},
month = {September},
year = {2024}
}
@article{qwen2,
title={Qwen2 Technical Report},
author={An Yang and Baosong Yang and Binyuan Hui and Bo Zheng and Bowen Yu and Chang Zhou and Chengpeng Li and Chengyuan Li and Dayiheng Liu and Fei Huang and Guanting Dong and Haoran Wei and Huan Lin and Jialong Tang and Jialin Wang and Jian Yang and Jianhong Tu and Jianwei Zhang and Jianxin Ma and Jin Xu and Jingren Zhou and Jinze Bai and Jinzheng He and Junyang Lin and Kai Dang and Keming Lu and Keqin Chen and Kexin Yang and Mei Li and Mingfeng Xue and Na Ni and Pei Zhang and Peng Wang and Ru Peng and Rui Men and Ruize Gao and Runji Lin and Shijie Wang and Shuai Bai and Sinan Tan and Tianhang Zhu and Tianhao Li and Tianyu Liu and Wenbin Ge and Xiaodong Deng and Xiaohuan Zhou and Xingzhang Ren and Xinyu Zhang and Xipin Wei and Xuancheng Ren and Yang Fan and Yang Yao and Yichang Zhang and Yu Wan and Yunfei Chu and Yuqiong Liu and Zeyu Cui and Zhenru Zhang and Zhihao Fan},
journal={arXiv preprint arXiv:2407.10671},
year={2024}
}
``` |
Mungert/Qwen2.5-VL-3B-Instruct-GGUF | Mungert | 2025-06-06T19:05:04 | 6,664 | 16 | transformers | [
"transformers",
"gguf",
"multimodal",
"image-text-to-text",
"en",
"arxiv:2309.00071",
"arxiv:2409.12191",
"arxiv:2308.12966",
"endpoints_compatible",
"region:us",
"imatrix",
"conversational"
] | image-text-to-text | 2025-03-27T23:20:23 |
---
license_name: qwen-research
license_link: https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct/blob/main/LICENSE
language:
- en
pipeline_tag: image-text-to-text
tags:
- multimodal
library_name: transformers
---
# <span style="color: #7FFF7F;">Qwen2.5-VL-3B-Instruct GGUF Models</span>
These files have been built using a imatrix file and latest llama.cpp build. You must use a fork of llama.cpp to use vision with the model.
## How to Use Qwen 2.5 VL Instruct with llama.cpp
To utilize the experimental support for Qwen 2.5 VL in `llama.cpp`, follow these steps:
Note this uses a fork of llama.cpp. At this time the main branch does not support vision for this model
1. **Clone the lastest llama.cpp Fork**:
```bash
git clone https://github.com/HimariO/llama.cpp.qwen2vl.git
cd llama.cpp.qwen2vl
git checkout qwen25-vl-20250404
```
2. **Build the Llama.cpp**:
Build llama.cpp as usual : https://github.com/ggml-org/llama.cpp#building-the-project
Once llama.cpp is built Copy the ./llama.cpp.qwen2vl/build/bin/llama-qwen2-vl-cli to a chosen folder.
3. **Download the Qwen 2.5 VL gguf file**:
https://huggingface.co/Mungert/Qwen2.5-VL-3B-Instruct-GGUF/tree/main
Choose a gguf file without the mmproj in the name
Example gguf file : https://huggingface.co/Mungert/Mungert/Qwen2.5-VL-3B-Instruct-GGUF/resolve/main/Qwen2.5-VL-3B-Instruct-q8_0.gguf
Copy this file to your chosen folder.
4. **Download the Qwen 2.5 VL mmproj file**
https://huggingface.co/Mungert/Qwen2.5-VL-3B-Instruct-GGUF/tree/main
Choose a file with mmproj in the name
Example mmproj file : https://huggingface.co/Mungert/Qwen2.5-VL-3B-Instruct-GGUF/resolve/main/Qwen2.5-VL-3B-Instruct-mmproj-f16.gguf
Copy this file to your chosen folder.
5. Copy images to the same folder as the gguf files or alter paths appropriately.
In the example below the gguf files, images and llama-qwen2vl-cli are in the same folder.
Example image: image https://huggingface.co/Mungert/Qwen2.5-VL-3B-Instruct-GGUF/resolve/main/car-1.jpg
Copy this file to your chosen folder.
6. **Run the CLI Tool**:
From your chosen folder :
```bash
llama-qwen2vl-cli -m Qwen2.5-VL-3B-Instruct-q8_0.gguf --mmproj Qwen2.5-VL-3B-Instruct-mmproj-f16.gguf -p "Describe this image." --image ./car-1.jpg
```
## <span style="color: #7FFF7F;">Ultra-Low-Bit Quantization with IQ-DynamicGate (1-2 bit)</span>
Our latest quantization method introduces **precision-adaptive quantization** for ultra-low-bit models (1-2 bit), with benchmark-proven improvements on **Llama-3-8B**. This approach uses layer-specific strategies to preserve accuracy while maintaining extreme memory efficiency.
### **Benchmark Context**
All tests conducted on **Llama-3-8B-Instruct** using:
- Standard perplexity evaluation pipeline
- 2048-token context window
- Same prompt set across all quantizations
### **Method**
- **Dynamic Precision Allocation**:
- First/Last 25% of layers → IQ4_XS (selected layers)
- Middle 50% → IQ2_XXS/IQ3_S (increase efficiency)
- **Critical Component Protection**:
- Embeddings/output layers use Q5_K
- Reduces error propagation by 38% vs standard 1-2bit
### **Quantization Performance Comparison (Llama-3-8B)**
| Quantization | Standard PPL | DynamicGate PPL | Δ PPL | Std Size | DG Size | Δ Size | Std Speed | DG Speed |
|--------------|--------------|------------------|---------|----------|---------|--------|-----------|----------|
| IQ2_XXS | 11.30 | 9.84 | -12.9% | 2.5G | 2.6G | +0.1G | 234s | 246s |
| IQ2_XS | 11.72 | 11.63 | -0.8% | 2.7G | 2.8G | +0.1G | 242s | 246s |
| IQ2_S | 14.31 | 9.02 | -36.9% | 2.7G | 2.9G | +0.2G | 238s | 244s |
| IQ1_M | 27.46 | 15.41 | -43.9% | 2.2G | 2.5G | +0.3G | 206s | 212s |
| IQ1_S | 53.07 | 32.00 | -39.7% | 2.1G | 2.4G | +0.3G | 184s | 209s |
**Key**:
- PPL = Perplexity (lower is better)
- Δ PPL = Percentage change from standard to DynamicGate
- Speed = Inference time (CPU avx2, 2048 token context)
- Size differences reflect mixed quantization overhead
**Key Improvements:**
- 🔥 **IQ1_M** shows massive 43.9% perplexity reduction (27.46 → 15.41)
- 🚀 **IQ2_S** cuts perplexity by 36.9% while adding only 0.2GB
- ⚡ **IQ1_S** maintains 39.7% better accuracy despite 1-bit quantization
**Tradeoffs:**
- All variants have modest size increases (0.1-0.3GB)
- Inference speeds remain comparable (<5% difference)
### **When to Use These Models**
📌 **Fitting models into GPU VRAM**
✔ **Memory-constrained deployments**
✔ **Cpu and Edge Devices** where 1-2bit errors can be tolerated
✔ **Research** into ultra-low-bit quantization
Quick test with highest of the DynamicGate quants IQ2_M on Qwen 2.5 VL 3B :
```bash
llama.cpp.qwen2vl/build/bin/llama-qwen2vl-cli -m Qwen2.5-VL-3B-Instruct-iq2_m.gguf --mmproj qwen.qwen2.5-vl-3b-instruct-vision.f16.gguf -p "Describe this image in a lot of detail." --image ./car-1.jpg
```
<p align="center">
<img src="https://huggingface.co/Mungert/Qwen2.5-VL-3B-Instruct-GGUF/resolve/main/car-1.jpg" width="80%"/>
<p>
Ouput :
The image depicts a sleek, black Porsche Panamera Turbo, captured in motion on what appears to be a racetrack or a high-speed road. The car is captured from a rear-side angle, showcasing its aerodynamic design and distinctive features. The vehicle's taillights are illuminated, creating a striking contrast with the dark body. The Porsche logo is prominently displayed on the rear, along with the words "Panamera Turbo" and the license plate "CVC-911." The license plate is accompanied by a California "COOPER" sticker, indicating the car might be registered in California. The road is blurred due to the speed, emphasizing the high performance and advanced engineering of the vehicle. The background features a mix of trees and a few streetlights, suggesting an evening or early evening setting.
**Wow that's impresive for a highly compresssed 3B model!**
With the lower quants the quality does suffer specially the xs quants. So if you need to squeeze the model into ram then give iq2_m a try.
## **Choosing the Right Model Format**
Selecting the correct model format depends on your **hardware capabilities** and **memory constraints**.
### **BF16 (Brain Float 16) – Use if BF16 acceleration is available**
- A 16-bit floating-point format designed for **faster computation** while retaining good precision.
- Provides **similar dynamic range** as FP32 but with **lower memory usage**.
- Recommended if your hardware supports **BF16 acceleration** (check your device’s specs).
- Ideal for **high-performance inference** with **reduced memory footprint** compared to FP32.
📌 **Use BF16 if:**
✔ Your hardware has native **BF16 support** (e.g., newer GPUs, TPUs).
✔ You want **higher precision** while saving memory.
✔ You plan to **requantize** the model into another format.
📌 **Avoid BF16 if:**
❌ Your hardware does **not** support BF16 (it may fall back to FP32 and run slower).
❌ You need compatibility with older devices that lack BF16 optimization.
---
### **F16 (Float 16) – More widely supported than BF16**
- A 16-bit floating-point **high precision** but with less of range of values than BF16.
- Works on most devices with **FP16 acceleration support** (including many GPUs and some CPUs).
- Slightly lower numerical precision than BF16 but generally sufficient for inference.
📌 **Use F16 if:**
✔ Your hardware supports **FP16** but **not BF16**.
✔ You need a **balance between speed, memory usage, and accuracy**.
✔ You are running on a **GPU** or another device optimized for FP16 computations.
📌 **Avoid F16 if:**
❌ Your device lacks **native FP16 support** (it may run slower than expected).
❌ You have memory limitations.
---
### **Quantized Models (Q4_K, Q6_K, Q8, etc.) – For CPU & Low-VRAM Inference**
Quantization reduces model size and memory usage while maintaining as much accuracy as possible.
- **Lower-bit models (Q4_K)** → **Best for minimal memory usage**, may have lower precision.
- **Higher-bit models (Q6_K, Q8_0)** → **Better accuracy**, requires more memory.
📌 **Use Quantized Models if:**
✔ You are running inference on a **CPU** and need an optimized model.
✔ Your device has **low VRAM** and cannot load full-precision models.
✔ You want to reduce **memory footprint** while keeping reasonable accuracy.
📌 **Avoid Quantized Models if:**
❌ You need **maximum accuracy** (full-precision models are better for this).
❌ Your hardware has enough VRAM for higher-precision formats (BF16/F16).
---
### **Very Low-Bit Quantization (IQ3_XS, IQ3_S, IQ3_M, Q4_K, Q4_0)**
These models are optimized for **extreme memory efficiency**, making them ideal for **low-power devices** or **large-scale deployments** where memory is a critical constraint.
- **IQ3_XS**: Ultra-low-bit quantization (3-bit) with **extreme memory efficiency**.
- **Use case**: Best for **ultra-low-memory devices** where even Q4_K is too large.
- **Trade-off**: Lower accuracy compared to higher-bit quantizations.
- **IQ3_S**: Small block size for **maximum memory efficiency**.
- **Use case**: Best for **low-memory devices** where **IQ3_XS** is too aggressive.
- **IQ3_M**: Medium block size for better accuracy than **IQ3_S**.
- **Use case**: Suitable for **low-memory devices** where **IQ3_S** is too limiting.
- **Q4_K**: 4-bit quantization with **block-wise optimization** for better accuracy.
- **Use case**: Best for **low-memory devices** where **Q6_K** is too large.
- **Q4_0**: Pure 4-bit quantization, optimized for **ARM devices**.
- **Use case**: Best for **ARM-based devices** or **low-memory environments**.
---
### **Summary Table: Model Format Selection**
| Model Format | Precision | Memory Usage | Device Requirements | Best Use Case |
|--------------|------------|---------------|----------------------|---------------|
| **BF16** | Highest | High | BF16-supported GPU/CPUs | High-speed inference with reduced memory |
| **F16** | High | High | FP16-supported devices | GPU inference when BF16 isn’t available |
| **Q4_K** | Medium Low | Low | CPU or Low-VRAM devices | Best for memory-constrained environments |
| **Q6_K** | Medium | Moderate | CPU with more memory | Better accuracy while still being quantized |
| **Q8_0** | High | Moderate | CPU or GPU with enough VRAM | Best accuracy among quantized models |
| **IQ3_XS** | Very Low | Very Low | Ultra-low-memory devices | Extreme memory efficiency and low accuracy |
| **Q4_0** | Low | Low | ARM or low-memory devices | llama.cpp can optimize for ARM devices |
---
## **Included Files & Details**
### `Qwen2.5-VL-3B-Instruct-bf16.gguf`
- Model weights preserved in **BF16**.
- Use this if you want to **requantize** the model into a different format.
- Best if your device supports **BF16 acceleration**.
### `Qwen2.5-VL-3B-Instruct-f16.gguf`
- Model weights stored in **F16**.
- Use if your device supports **FP16**, especially if BF16 is not available.
### `Qwen2.5-VL-3B-Instruct-bf16-q8_0.gguf`
- **Output & embeddings** remain in **BF16**.
- All other layers quantized to **Q8_0**.
- Use if your device supports **BF16** and you want a quantized version.
### `Qwen2.5-VL-3B-Instruct-f16-q8_0.gguf`
- **Output & embeddings** remain in **F16**.
- All other layers quantized to **Q8_0**.
### `Qwen2.5-VL-3B-Instruct-q4_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q4_K**.
- Good for **CPU inference** with limited memory.
### `Qwen2.5-VL-3B-Instruct-q4_k_s.gguf`
- Smallest **Q4_K** variant, using less memory at the cost of accuracy.
- Best for **very low-memory setups**.
### `Qwen2.5-VL-3B-Instruct-q6_k.gguf`
- **Output & embeddings** quantized to **Q8_0**.
- All other layers quantized to **Q6_K** .
### `Qwen2.5-VL-3B-Instruct-q8_0.gguf`
- Fully **Q8** quantized model for better accuracy.
- Requires **more memory** but offers higher precision.
### `Qwen2.5-VL-3B-Instruct-iq3_xs.gguf`
- **IQ3_XS** quantization, optimized for **extreme memory efficiency**.
- Best for **ultra-low-memory devices**.
### `Qwen2.5-VL-3B-Instruct-iq3_m.gguf`
- **IQ3_M** quantization, offering a **medium block size** for better accuracy.
- Suitable for **low-memory devices**.
### `Qwen2.5-VL-3B-Instruct-q4_0.gguf`
- Pure **Q4_0** quantization, optimized for **ARM devices**.
- Best for **low-memory environments**.
- Prefer IQ4_NL for better accuracy.
# <span id="testllm" style="color: #7F7FFF;">🚀 If you find these models useful</span>
Please click like ❤ . Also I’d really appreciate it if you could test my Network Monitor Assistant at 👉 [Network Monitor Assitant](https://readyforquantum.com).
💬 Click the **chat icon** (bottom right of the main and dashboard pages) . Choose a LLM; toggle between the LLM Types TurboLLM -> FreeLLM -> TestLLM.
### What I'm Testing
I'm experimenting with **function calling** against my network monitoring service. Using small open source models. I am into the question "How small can it go and still function".
🟡 **TestLLM** – Runs the current testing model using llama.cpp on 6 threads of a Cpu VM (Should take about 15s to load. Inference speed is quite slow and it only processes one user prompt at a time—still working on scaling!). If you're curious, I'd be happy to share how it works! .
### The other Available AI Assistants
🟢 **TurboLLM** – Uses **gpt-4o-mini** Fast! . Note: tokens are limited since OpenAI models are pricey, but you can [Login](https://readyforquantum.com) or [Download](https://readyforquantum.com/download/?utm_source=huggingface&utm_medium=referral&utm_campaign=huggingface_repo_readme) the Free Network Monitor agent to get more tokens, Alternatively use the TestLLM .
🔵 **HugLLM** – Runs **open-source Hugging Face models** Fast, Runs small models (≈8B) hence lower quality, Get 2x more tokens (subject to Hugging Face API availability)
# Qwen2.5-VL-3B-Instruct
<a href="https://chat.qwenlm.ai/" target="_blank" style="margin: 2px;">
<img alt="Chat" src="https://img.shields.io/badge/%F0%9F%92%9C%EF%B8%8F%20Qwen%20Chat%20-536af5" style="display: inline-block; vertical-align: middle;"/>
</a>
## Introduction
In the past five months since Qwen2-VL’s release, numerous developers have built new models on the Qwen2-VL vision-language models, providing us with valuable feedback. During this period, we focused on building more useful vision-language models. Today, we are excited to introduce the latest addition to the Qwen family: Qwen2.5-VL.
#### Key Enhancements:
* **Understand things visually**: Qwen2.5-VL is not only proficient in recognizing common objects such as flowers, birds, fish, and insects, but it is highly capable of analyzing texts, charts, icons, graphics, and layouts within images.
* **Being agentic**: Qwen2.5-VL directly plays as a visual agent that can reason and dynamically direct tools, which is capable of computer use and phone use.
* **Understanding long videos and capturing events**: Qwen2.5-VL can comprehend videos of over 1 hour, and this time it has a new ability of cpaturing event by pinpointing the relevant video segments.
* **Capable of visual localization in different formats**: Qwen2.5-VL can accurately localize objects in an image by generating bounding boxes or points, and it can provide stable JSON outputs for coordinates and attributes.
* **Generating structured outputs**: for data like scans of invoices, forms, tables, etc. Qwen2.5-VL supports structured outputs of their contents, benefiting usages in finance, commerce, etc.
#### Model Architecture Updates:
* **Dynamic Resolution and Frame Rate Training for Video Understanding**:
We extend dynamic resolution to the temporal dimension by adopting dynamic FPS sampling, enabling the model to comprehend videos at various sampling rates. Accordingly, we update mRoPE in the time dimension with IDs and absolute time alignment, enabling the model to learn temporal sequence and speed, and ultimately acquire the ability to pinpoint specific moments.
<p align="center">
<img src="https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2.5-VL/qwen2.5vl_arc.jpeg" width="80%"/>
<p>
* **Streamlined and Efficient Vision Encoder**
We enhance both training and inference speeds by strategically implementing window attention into the ViT. The ViT architecture is further optimized with SwiGLU and RMSNorm, aligning it with the structure of the Qwen2.5 LLM.
We have three models with 3, 7 and 72 billion parameters. This repo contains the instruction-tuned 3B Qwen2.5-VL model. For more information, visit our [Blog](https://qwenlm.github.io/blog/qwen2.5-vl/) and [GitHub](https://github.com/QwenLM/Qwen2.5-VL).
## Evaluation
### Image benchmark
| Benchmark | InternVL2.5-4B |Qwen2-VL-7B |Qwen2.5-VL-3B |
| :--- | :---: | :---: | :---: |
| MMMU<sub>val</sub> | 52.3 | 54.1 | 53.1|
| MMMU-Pro<sub>val</sub> | **32.7** | 30.5 | 31.6|
| AI2D<sub>test</sub> | 81.4 | **83.0** | 81.5 |
| DocVQA<sub>test</sub> | 91.6 | 94.5 | **93.9** |
| InfoVQA<sub>test</sub> | 72.1 | 76.5 | **77.1** |
| TextVQA<sub>val</sub> | 76.8 | **84.3** | 79.3|
| MMBench-V1.1<sub>test</sub> | 79.3 | **80.7** | 77.6 |
| MMStar | 58.3 | **60.7** | 55.9 |
| MathVista<sub>testmini</sub> | 60.5 | 58.2 | **62.3** |
| MathVision<sub>full</sub> | 20.9 | 16.3 | **21.2** |
### Video benchmark
| Benchmark | InternVL2.5-4B | Qwen2-VL-7B | Qwen2.5-VL-3B |
| :--- | :---: | :---: | :---: |
| MVBench | 71.6 | 67.0 | 67.0 |
| VideoMME | 63.6/62.3 | 69.0/63.3 | 67.6/61.5 |
| MLVU | 48.3 | - | 68.2 |
| LVBench | - | - | 43.3 |
| MMBench-Video | 1.73 | 1.44 | 1.63 |
| EgoSchema | - | - | 64.8 |
| PerceptionTest | - | - | 66.9 |
| TempCompass | - | - | 64.4 |
| LongVideoBench | 55.2 | 55.6 | 54.2 |
| CharadesSTA/mIoU | - | - | 38.8 |
### Agent benchmark
| Benchmarks | Qwen2.5-VL-3B |
|-------------------------|---------------|
| ScreenSpot | 55.5 |
| ScreenSpot Pro | 23.9 |
| AITZ_EM | 76.9 |
| Android Control High_EM | 63.7 |
| Android Control Low_EM | 22.2 |
| AndroidWorld_SR | 90.8 |
| MobileMiniWob++_SR | 67.9 |
## Requirements
The code of Qwen2.5-VL has been in the latest Hugging face transformers and we advise you to build from source with command:
```
pip install git+https://github.com/huggingface/transformers accelerate
```
or you might encounter the following error:
```
KeyError: 'qwen2_5_vl'
```
## Quickstart
Below, we provide simple examples to show how to use Qwen2.5-VL with 🤖 ModelScope and 🤗 Transformers.
The code of Qwen2.5-VL has been in the latest Hugging face transformers and we advise you to build from source with command:
```
pip install git+https://github.com/huggingface/transformers accelerate
```
or you might encounter the following error:
```
KeyError: 'qwen2_5_vl'
```
We offer a toolkit to help you handle various types of visual input more conveniently, as if you were using an API. This includes base64, URLs, and interleaved images and videos. You can install it using the following command:
```bash
# It's highly recommanded to use `[decord]` feature for faster video loading.
pip install qwen-vl-utils[decord]==0.0.8
```
If you are not using Linux, you might not be able to install `decord` from PyPI. In that case, you can use `pip install qwen-vl-utils` which will fall back to using torchvision for video processing. However, you can still [install decord from source](https://github.com/dmlc/decord?tab=readme-ov-file#install-from-source) to get decord used when loading video.
### Using 🤗 Transformers to Chat
Here we show a code snippet to show you how to use the chat model with `transformers` and `qwen_vl_utils`:
```python
from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info
# default: Load the model on the available device(s)
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
"Qwen/Qwen2.5-VL-3B-Instruct", torch_dtype="auto", device_map="auto"
)
# We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
# model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
# "Qwen/Qwen2.5-VL-3B-Instruct",
# torch_dtype=torch.bfloat16,
# attn_implementation="flash_attention_2",
# device_map="auto",
# )
# default processer
processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-3B-Instruct")
# The default range for the number of visual tokens per image in the model is 4-16384.
# You can set min_pixels and max_pixels according to your needs, such as a token range of 256-1280, to balance performance and cost.
# min_pixels = 256*28*28
# max_pixels = 1280*28*28
# processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-3B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels)
messages = [
{
"role": "user",
"content": [
{
"type": "image",
"image": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg",
},
{"type": "text", "text": "Describe this image."},
],
}
]
# Preparation for inference
text = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
image_inputs, video_inputs = process_vision_info(messages)
inputs = processor(
text=[text],
images=image_inputs,
videos=video_inputs,
padding=True,
return_tensors="pt",
)
inputs = inputs.to("cuda")
# Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=128)
generated_ids_trimmed = [
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)
```
<details>
<summary>Multi image inference</summary>
```python
# Messages containing multiple images and a text query
messages = [
{
"role": "user",
"content": [
{"type": "image", "image": "file:///path/to/image1.jpg"},
{"type": "image", "image": "file:///path/to/image2.jpg"},
{"type": "text", "text": "Identify the similarities between these images."},
],
}
]
# Preparation for inference
text = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
image_inputs, video_inputs = process_vision_info(messages)
inputs = processor(
text=[text],
images=image_inputs,
videos=video_inputs,
padding=True,
return_tensors="pt",
)
inputs = inputs.to("cuda")
# Inference
generated_ids = model.generate(**inputs, max_new_tokens=128)
generated_ids_trimmed = [
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)
```
</details>
<details>
<summary>Video inference</summary>
```python
# Messages containing a images list as a video and a text query
messages = [
{
"role": "user",
"content": [
{
"type": "video",
"video": [
"file:///path/to/frame1.jpg",
"file:///path/to/frame2.jpg",
"file:///path/to/frame3.jpg",
"file:///path/to/frame4.jpg",
],
},
{"type": "text", "text": "Describe this video."},
],
}
]
# Messages containing a local video path and a text query
messages = [
{
"role": "user",
"content": [
{
"type": "video",
"video": "file:///path/to/video1.mp4",
"max_pixels": 360 * 420,
"fps": 1.0,
},
{"type": "text", "text": "Describe this video."},
],
}
]
# Messages containing a video url and a text query
messages = [
{
"role": "user",
"content": [
{
"type": "video",
"video": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2-VL/space_woaudio.mp4",
},
{"type": "text", "text": "Describe this video."},
],
}
]
#In Qwen 2.5 VL, frame rate information is also input into the model to align with absolute time.
# Preparation for inference
text = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
image_inputs, video_inputs, video_kwargs = process_vision_info(messages, return_video_kwargs=True)
inputs = processor(
text=[text],
images=image_inputs,
videos=video_inputs,
fps=fps,
padding=True,
return_tensors="pt",
**video_kwargs,
)
inputs = inputs.to("cuda")
# Inference
generated_ids = model.generate(**inputs, max_new_tokens=128)
generated_ids_trimmed = [
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)
```
Video URL compatibility largely depends on the third-party library version. The details are in the table below. change the backend by `FORCE_QWENVL_VIDEO_READER=torchvision` or `FORCE_QWENVL_VIDEO_READER=decord` if you prefer not to use the default one.
| Backend | HTTP | HTTPS |
|-------------|------|-------|
| torchvision >= 0.19.0 | ✅ | ✅ |
| torchvision < 0.19.0 | ❌ | ❌ |
| decord | ✅ | ❌ |
</details>
<details>
<summary>Batch inference</summary>
```python
# Sample messages for batch inference
messages1 = [
{
"role": "user",
"content": [
{"type": "image", "image": "file:///path/to/image1.jpg"},
{"type": "image", "image": "file:///path/to/image2.jpg"},
{"type": "text", "text": "What are the common elements in these pictures?"},
],
}
]
messages2 = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Who are you?"},
]
# Combine messages for batch processing
messages = [messages1, messages2]
# Preparation for batch inference
texts = [
processor.apply_chat_template(msg, tokenize=False, add_generation_prompt=True)
for msg in messages
]
image_inputs, video_inputs = process_vision_info(messages)
inputs = processor(
text=texts,
images=image_inputs,
videos=video_inputs,
padding=True,
return_tensors="pt",
)
inputs = inputs.to("cuda")
# Batch Inference
generated_ids = model.generate(**inputs, max_new_tokens=128)
generated_ids_trimmed = [
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_texts = processor.batch_decode(
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_texts)
```
</details>
### 🤖 ModelScope
We strongly advise users especially those in mainland China to use ModelScope. `snapshot_download` can help you solve issues concerning downloading checkpoints.
### More Usage Tips
For input images, we support local files, base64, and URLs. For videos, we currently only support local files.
```python
# You can directly insert a local file path, a URL, or a base64-encoded image into the position where you want in the text.
## Local file path
messages = [
{
"role": "user",
"content": [
{"type": "image", "image": "file:///path/to/your/image.jpg"},
{"type": "text", "text": "Describe this image."},
],
}
]
## Image URL
messages = [
{
"role": "user",
"content": [
{"type": "image", "image": "http://path/to/your/image.jpg"},
{"type": "text", "text": "Describe this image."},
],
}
]
## Base64 encoded image
messages = [
{
"role": "user",
"content": [
{"type": "image", "image": "data:image;base64,/9j/..."},
{"type": "text", "text": "Describe this image."},
],
}
]
```
#### Image Resolution for performance boost
The model supports a wide range of resolution inputs. By default, it uses the native resolution for input, but higher resolutions can enhance performance at the cost of more computation. Users can set the minimum and maximum number of pixels to achieve an optimal configuration for their needs, such as a token count range of 256-1280, to balance speed and memory usage.
```python
min_pixels = 256 * 28 * 28
max_pixels = 1280 * 28 * 28
processor = AutoProcessor.from_pretrained(
"Qwen/Qwen2.5-VL-3B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels
)
```
Besides, We provide two methods for fine-grained control over the image size input to the model:
1. Define min_pixels and max_pixels: Images will be resized to maintain their aspect ratio within the range of min_pixels and max_pixels.
2. Specify exact dimensions: Directly set `resized_height` and `resized_width`. These values will be rounded to the nearest multiple of 28.
```python
# min_pixels and max_pixels
messages = [
{
"role": "user",
"content": [
{
"type": "image",
"image": "file:///path/to/your/image.jpg",
"resized_height": 280,
"resized_width": 420,
},
{"type": "text", "text": "Describe this image."},
],
}
]
# resized_height and resized_width
messages = [
{
"role": "user",
"content": [
{
"type": "image",
"image": "file:///path/to/your/image.jpg",
"min_pixels": 50176,
"max_pixels": 50176,
},
{"type": "text", "text": "Describe this image."},
],
}
]
```
### Processing Long Texts
The current `config.json` is set for context length up to 32,768 tokens.
To handle extensive inputs exceeding 32,768 tokens, we utilize [YaRN](https://arxiv.org/abs/2309.00071), a technique for enhancing model length extrapolation, ensuring optimal performance on lengthy texts.
For supported frameworks, you could add the following to `config.json` to enable YaRN:
```
{
...,
"type": "yarn",
"mrope_section": [
16,
24,
24
],
"factor": 4,
"original_max_position_embeddings": 32768
}
```
However, it should be noted that this method has a significant impact on the performance of temporal and spatial localization tasks, and is therefore not recommended for use.
At the same time, for long video inputs, since MRoPE itself is more economical with ids, the max_position_embeddings can be directly modified to a larger value, such as 64k.
## Citation
If you find our work helpful, feel free to give us a cite.
```
@misc{qwen2.5-VL,
title = {Qwen2.5-VL},
url = {https://qwenlm.github.io/blog/qwen2.5-vl/},
author = {Qwen Team},
month = {January},
year = {2025}
}
@article{Qwen2VL,
title={Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution},
author={Wang, Peng and Bai, Shuai and Tan, Sinan and Wang, Shijie and Fan, Zhihao and Bai, Jinze and Chen, Keqin and Liu, Xuejing and Wang, Jialin and Ge, Wenbin and Fan, Yang and Dang, Kai and Du, Mengfei and Ren, Xuancheng and Men, Rui and Liu, Dayiheng and Zhou, Chang and Zhou, Jingren and Lin, Junyang},
journal={arXiv preprint arXiv:2409.12191},
year={2024}
}
@article{Qwen-VL,
title={Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond},
author={Bai, Jinze and Bai, Shuai and Yang, Shusheng and Wang, Shijie and Tan, Sinan and Wang, Peng and Lin, Junyang and Zhou, Chang and Zhou, Jingren},
journal={arXiv preprint arXiv:2308.12966},
year={2023}
}
```
|
andrewsamce/ppo-LunarLander-v2 | andrewsamce | 2025-06-06T19:01:48 | 0 | 0 | stable-baselines3 | [
"stable-baselines3",
"LunarLander-v2",
"deep-reinforcement-learning",
"reinforcement-learning",
"model-index",
"region:us"
] | reinforcement-learning | 2025-06-06T19:01:27 | ---
library_name: stable-baselines3
tags:
- LunarLander-v2
- deep-reinforcement-learning
- reinforcement-learning
- stable-baselines3
model-index:
- name: PPO
results:
- task:
type: reinforcement-learning
name: reinforcement-learning
dataset:
name: LunarLander-v2
type: LunarLander-v2
metrics:
- type: mean_reward
value: 268.76 +/- 15.19
name: mean_reward
verified: false
---
# **PPO** Agent playing **LunarLander-v2**
This is a trained model of a **PPO** agent playing **LunarLander-v2**
using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3).
## Usage (with Stable-baselines3)
TODO: Add your code
```python
from stable_baselines3 import ...
from huggingface_sb3 import load_from_hub
...
```
|
End of preview. Expand
in Data Studio
Dataset Card for Hugging Face Hub Model Cards
This datasets consists of model cards for models hosted on the Hugging Face Hub. The model cards are created by the community and provide information about the model, its performance, its intended uses, and more. This dataset is updated on a daily basis and includes publicly available models on the Hugging Face Hub.
This dataset is made available to help support users wanting to work with a large number of Model Cards from the Hub. We hope that this dataset will help support research in the area of Model Cards and their use but the format of this dataset may not be useful for all use cases. If there are other features that you would like to see included in this dataset, please open a new discussion.
Dataset Details
Uses
There are a number of potential uses for this dataset including:
- text mining to find common themes in model cards
- analysis of the model card format/content
- topic modelling of model cards
- analysis of the model card metadata
- training language models on model cards
Out-of-Scope Use
[More Information Needed]
Dataset Structure
This dataset has a single split.
Dataset Creation
Curation Rationale
The dataset was created to assist people in working with model cards. In particular it was created to support research in the area of model cards and their use. It is possible to use the Hugging Face Hub API or client library to download model cards and this option may be preferable if you have a very specific use case or require a different format.
Source Data
The source data is README.md files for models hosted on the Hugging Face Hub. We do not include any other supplementary files that may be included in the model card directory.
Data Collection and Processing
The data is downloaded using a CRON job on a daily basis.
Who are the source data producers?
The source data producers are the creators of the model cards on the Hugging Face Hub. This includes a broad variety of people from the community ranging from large companies to individual researchers. We do not gather any information about who created the model card in this repository although this information can be gathered from the Hugging Face Hub API.
Annotations [optional]
There are no additional annotations in this dataset beyond the model card content.
Annotation process
N/A
Who are the annotators?
N/A
Personal and Sensitive Information
We make no effort to anonymize the data. Whilst we don't expect the majority of model cards to contain personal or sensitive information, it is possible that some model cards may contain this information. Model cards may also link to websites or email addresses.
Bias, Risks, and Limitations
Model cards are created by the community and we do not have any control over the content of the model cards. We do not review the content of the model cards and we do not make any claims about the accuracy of the information in the model cards. Some model cards will themselves discuss bias and sometimes this is done by providing examples of bias in either the training data or the responses provided by the model. As a result this dataset may contain examples of bias.
Whilst we do not directly download any images linked to in the model cards, some model cards may include images. Some of these images may not be suitable for all audiences.
Recommendations
Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.
Citation
No formal citation is required for this dataset but if you use this dataset in your work, please include a link to this dataset page.
Dataset Card Authors
Dataset Card Contact
- Downloads last month
- 1,360
Curated by:
Daniel van Strien
Language(s) (NLP):
Model cards on the Hugging Face Hub are predominantly in English but may include other languages.
Size of downloaded dataset files:
917 MB
Size of the auto-converted Parquet files:
917 MB
Number of rows:
803,915