--- language: - en library_name: transformers license: apache-2.0 metrics: - accuracy tags: - multimodal pipeline_tag: video-text-to-text --- # Video-XL-2 [\[📰 Blog\]](https://unabletousegit.github.io/video-xl2.github.io/) [\[📂 GitHub\]](https://github.com/VectorSpaceLab/Video-XL) [\[📜 Tech Report(comming soon)\]]() ## How to use the model Video-XL-2 supply two efficiency optimization strategy: chunk-based prefill and bi-level kvs decoding. You can flexibly choose them based on your needs. TODO - [X] Release model weights. - [X] Release the inference code w/o. efficiency optimization. - [X] Release the inference code w. chunk-based prefill. - [ ] Release the inference code w. chunk-based prefill & bi-level kvs decoding. **Tips: Our inference code still under updating, you could update it by assign "--include '\*.py'" in huggingface-cli to only update the inference code, avoid downloading the whole model.* --- ### 0. Installing Required Packages ```bash pip install transformers==4.43.0 pip install torch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 --index-url https://download.pytorch.org/whl/cu121 pip install decord pip install einops pip install opencv-python pip install accelerate==0.30.0 pip install numpy==1.26.4 # optional pip install flash-attn --no-build-isolation ``` --- ### 1. Inference w/o. Efficiency Optimization ```python from transformers import AutoTokenizer, AutoModel, AutoConfig, BitsAndBytesConfig, AutoModelForCausalLM import torch # load model model_path = '/root/Models/Video-XL-2' tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) device = 'cuda' if torch.cuda.is_available() else 'cpu' model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True, device_map=device,quantization_config=None, attn_implementation="sdpa", torch_dtype=torch.float16, low_cpu_mem_usage=True) gen_kwargs = { "do_sample": False, "temperature": 0.01, "top_p": 0.001, "num_beams": 1, "use_cache": True, "max_new_tokens": 256 } model.config.enable_sparse = False # input data video_path = "/asset/demo.mp4" question1 = "How many people in the video? (A)3 people (B)6 people. Please only respone the letter" # params max_num_frames = 150 sample_fps = 1 # extract frame at 1fps max_sample_fps = 4 with torch.inference_mode(): response = model.chat(video_path, tokenizer, question1, chat_history=None, return_history=False,max_num_frames=max_num_frames, sample_fps=sample_fps, max_sample_fps=max_sample_fps, generation_config=gen_kwargs) print(response) ``` --- ### 2. Inference w. Chunk-based Pre-filling Chunk-based prefill significantly reduces memory demands and response latency by encoding video input in a streaming manner. This advantage becomes particularly noticeable with longer videos. To enable this mode, you need to set `enable_chunk_prefill` to `True` and configure the `prefill_config` parameters: * **`chunk_prefill_mode`**: This defines the mode of chunk-based prefill. We currently support two modes: * **`streaming`**: This mode encodes video chunks streamingly. * **`mask`**: This mode achieves an equivalent effect using an attention mask. However, due to a lack of underlying optimized operators, the `mask` mode doesn't offer any efficiency improvements at this time. We recommend using the `streaming` mode. * **`chunk_size`**: This parameter specifies the size of each chunk processed in a single forward pass. The unit for `chunk_size` is **4 frames** (e.g., `chunk_size = 4` means processing visual tokens from **4×4 = 16 frames** at once). A larger `chunk_size` will gradually approach full attention, resulting in a higher peak memory usage. * **`step_size`**: This controls the step size between chunks. A smaller `step_size` leads to more continuous information transfer between chunks but may slightly decrease inference speed. * **`offload`**: This boolean parameter determines whether to offload the key-value states (KVs) of each chunk to the CPU during forwarding. While this can reduce memory usage, it will also lower the inference speed. * **`chunk_size_for_vision_tower`**: For longer video inputs, the vision tower can become a memory bottleneck during forwarding. To mitigate this, we also support a streaming mode for the vision tower, which is controlled by this parameter. The unit for `chunk_size_for_vision_tower` is **1 frames**. And, the value of `chunk_size_for_vision_tower` must be **a multiple of 4**. **Tip: Currently, chunk-based prefill only supports the 'sdpa' attention implementation.* ```python from transformers import AutoTokenizer, AutoModel, AutoConfig, BitsAndBytesConfig, AutoModelForCausalLM import torch import pdb import argparse torch.cuda.reset_peak_memory_stats() # load model model_path = '/root/Models/Video-XL-2' tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) device = 'cuda:0' if torch.cuda.is_available() else 'cpu' model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True, device_map=device,quantization_config=None, attn_implementation="sdpa", torch_dtype=torch.float16, low_cpu_mem_usage=True) # sdpa gen_kwargs = {"do_sample": False, "temperature": 0.01, "top_p": 0.001, "num_beams": 1, "use_cache": True, "max_new_tokens": 128} model.config.enable_chunk_prefill = True prefill_config = { 'chunk_prefill_mode': 'streaming', 'chunk_size': 4, 'step_size': 1, 'offload': True, 'chunk_size_for_vision_tower': 24, } model.config.prefill_config = prefill_config # input data video_path = "/asset/demo.mp4" question1 = "How many people in the video? (A)3 people (B)6 people. Please only respone the letter" # params max_num_frames = 1300 sample_fps = None # uniform sampling max_sample_fps = None with torch.inference_mode(): response = model.chat(video_path, tokenizer, question1, chat_history=None, return_history=False,max_num_frames=max_num_frames, sample_fps=sample_fps, max_sample_fps=max_sample_fps, generation_config=gen_kwargs) peak_memory_allocated = torch.cuda.max_memory_allocated() print(f"Memory Peak: {peak_memory_allocated / (1024**3):.2f} GB") print(response) ``` --- ### 3. Inference w. Chunk-based Pre-filling & Bi-level KVs Decoding coming soon ```python ``` ## ✏️ Citation ```bibtex @article{shu2024video, title={Video-XL: Extra-Long Vision Language Model for Hour-Scale Video Understanding}, author={Shu, Yan and Zhang, Peitian and Liu, Zheng and Qin, Minghao and Zhou, Junjie and Huang, Tiejun and Zhao, Bo}, journal={arXiv preprint arXiv:2409.14485}, year={2024} } @article{liu2025video, title={Video-XL-Pro: Reconstructive Token Compression for Extremely Long Video Understanding}, author={Liu, Xiangrui and Shu, Yan and Liu, Zheng and Li, Ao and Tian, Yang and Zhao, Bo}, journal={arXiv preprint arXiv:2503.18478}, year={2025} } ```