Quantized version of Mistral-Small-24B-Instruct-2501.
This model was obtained by quantizing the weights to INT4 data type, ready for inference with vLLM. This optimization reduces the number of bits per parameter from 16 to 4, reducing the disk size and GPU memory requirements by approximately 75%. Only the weights of the linear operators within transformers blocks are quantized.
This model can be deployed efficiently using the vLLM backend, as shown in the example below.
from transformers import AutoTokenizer
from vllm import LLM, SamplingParams
max_model_len, tp_size = 4096, 1
model_name = "neuralmagic/Mistral-Small-24B-Instruct-2501-quantized.w4a16"
tokenizer = AutoTokenizer.from_pretrained(model_name)
llm = LLM(model=model_name, tensor_parallel_size=tp_size, max_model_len=max_model_len, trust_remote_code=True)
sampling_params = SamplingParams(temperature=0.3, max_tokens=256, stop_token_ids=[tokenizer.eos_token_id])
messages_list = [
[{"role": "user", "content": "Who are you? Please respond in pirate speak!"}],
]
prompt_token_ids = [tokenizer.apply_chat_template(messages, add_generation_prompt=True) for messages in messages_list]
outputs = llm.generate(prompt_token_ids=prompt_token_ids, sampling_params=sampling_params)
generated_text = [output.outputs[0].text for output in outputs]
print(generated_text)
vLLM also supports OpenAI-compatible serving. See the documentation for more details.
podman run --rm -it --device nvidia.com/gpu=all -p 8000:8000 \
--ipc=host \
--env "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" \
--env "HF_HUB_OFFLINE=0" -v ~/.cache/vllm:/home/vllm/.cache \
--name=vllm \
registry.access.redhat.com/rhaiis/rh-vllm-cuda \
vllm serve \
--tensor-parallel-size 8 \
--max-model-len 32768 \
--enforce-eager --model RedHatAI/Mistral-Small-24B-Instruct-2501-quantized.w4a16
See Red Hat AI Inference Server documentation for more details.
# Download model from Red Hat Registry via docker
# Note: This downloads the model to ~/.cache/instructlab/models unless --model-dir is specified.
ilab model download --repository docker://registry.redhat.io/rhelai1/mistral-small-24b-instruct-2501-quantized-w4a16:1.5
# Serve model via ilab
ilab model serve --model-path ~/.cache/instructlab/models/mistral-small-24b-instruct-2501-quantized-w4a16 --gpu 1 -- --trust-remote-code
# Chat with model
ilab model chat --model ~/.cache/instructlab/models/mistral-small-24b-instruct-2501-quantized-w4a16
See Red Hat Enterprise Linux AI documentation for more details.
# Setting up vllm server with ServingRuntime
# Save as: vllm-servingruntime.yaml
apiVersion: serving.kserve.io/v1alpha1
kind: ServingRuntime
metadata:
name: vllm-cuda-runtime # OPTIONAL CHANGE: set a unique name
annotations:
openshift.io/display-name: vLLM NVIDIA GPU ServingRuntime for KServe
opendatahub.io/recommended-accelerators: '["nvidia.com/gpu"]'
labels:
opendatahub.io/dashboard: 'true'
spec:
annotations:
prometheus.io/port: '8080'
prometheus.io/path: '/metrics'
multiModel: false
supportedModelFormats:
- autoSelect: true
name: vLLM
containers:
- name: kserve-container
image: quay.io/modh/vllm:rhoai-2.20-cuda # CHANGE if needed. If AMD: quay.io/modh/vllm:rhoai-2.20-rocm
command:
- python
- -m
- vllm.entrypoints.openai.api_server
args:
- "--port=8080"
- "--model=/mnt/models"
- "--served-model-name={{.Name}}"
env:
- name: HF_HOME
value: /tmp/hf_home
ports:
- containerPort: 8080
protocol: TCP
# Attach model to vllm server. This is an NVIDIA template
# Save as: inferenceservice.yaml
apiVersion: serving.kserve.io/v1beta1
kind: InferenceService
metadata:
annotations:
openshift.io/display-name: Mistral-Small-24B-Instruct-2501-quantized.w4a16 # OPTIONAL CHANGE
serving.kserve.io/deploymentMode: RawDeployment
name: Mistral-Small-24B-Instruct-2501-quantized.w4a16 # specify model name. This value will be used to invoke the model in the payload
labels:
opendatahub.io/dashboard: 'true'
spec:
predictor:
maxReplicas: 1
minReplicas: 1
model:
args:
- "--trust-remote-code"
modelFormat:
name: vLLM
name: ''
resources:
limits:
cpu: '2' # this is model specific
memory: 8Gi # this is model specific
nvidia.com/gpu: '1' # this is accelerator specific
requests: # same comment for this block
cpu: '1'
memory: 4Gi
nvidia.com/gpu: '1'
runtime: vllm-cuda-runtime # must match the ServingRuntime name above
storageUri: oci://registry.redhat.io/rhelai1/modelcar-mistral-small-24b-instruct-2501-quantized-w4a16:1.5
tolerations:
- effect: NoSchedule
key: nvidia.com/gpu
operator: Exists
# make sure first to be in the project where you want to deploy the model
# oc project <project-name>
# apply both resources to run model
# Apply the ServingRuntime
oc apply -f vllm-servingruntime.yaml
# Apply the InferenceService
oc apply -f qwen-inferenceservice.yaml
# Replace <inference-service-name> and <cluster-ingress-domain> below:
# - Run `oc get inferenceservice` to find your URL if unsure.
# Call the server using curl:
curl https://<inference-service-name>-predictor-default.<domain>/v1/chat/completions
-H "Content-Type: application/json" \
-d '{
"model": "Mistral-Small-24B-Instruct-2501-quantized.w4a16",
"stream": true,
"stream_options": {
"include_usage": true
},
"max_tokens": 1,
"messages": [
{
"role": "user",
"content": "How can a bee fly when its wings are so small?"
}
]
}'
See Red Hat Openshift AI documentation for more details.
This model was created with llm-compressor by running the code snippet below.
python quantize.py --model_path mistralai/Mistral-Small-24B-Instruct-2501 --quant_path "output_dir" --calib_size 1024 --dampening_frac 0.05 --observer minmax --actorder false
from datasets import load_dataset
from transformers import AutoTokenizer
from llmcompressor.modifiers.quantization import GPTQModifier
from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot, apply
import argparse
from compressed_tensors.quantization import QuantizationScheme, QuantizationArgs, QuantizationType, QuantizationStrategy
def parse_actorder(value):
# Interpret the input value for --actorder
if value.lower() == "false":
return False
elif value.lower() == "group":
return "group"
elif value.lower() == "weight":
return "weight"
else:
raise argparse.ArgumentTypeError("Invalid value for --actorder. Use 'group' or 'False'.")
parser = argparse.ArgumentParser()
parser.add_argument('--model_path', type=str)
parser.add_argument('--quant_path', type=str)
parser.add_argument('--num_bits', type=int, default=4)
parser.add_argument('--sequential_update', type=bool, default=True)
parser.add_argument('--calib_size', type=int, default=256)
parser.add_argument('--dampening_frac', type=float, default=0.05)
parser.add_argument('--observer', type=str, default="minmax")
parser.add_argument(
'--actorder',
type=parse_actorder,
default=False, # Default value is False
help="Specify actorder as 'group' (string) or False (boolean)."
)
args = parser.parse_args()
model = SparseAutoModelForCausalLM.from_pretrained(
args.model_path,
device_map="auto",
torch_dtype="auto",
use_cache=False,
)
tokenizer = AutoTokenizer.from_pretrained(args.model_path)
NUM_CALIBRATION_SAMPLES = args.calib_size
DATASET_ID = "garage-bAInd/Open-Platypus"
DATASET_SPLIT = "train"
ds = load_dataset(DATASET_ID, split=DATASET_SPLIT)
ds = ds.shuffle(seed=42).select(range(NUM_CALIBRATION_SAMPLES))
def preprocess(example):
concat_txt = example["instruction"] + "\n" + example["output"]
return {"text": concat_txt}
ds = ds.map(preprocess)
def tokenize(sample):
return tokenizer(
sample["text"],
padding=False,
truncation=False,
add_special_tokens=True,
)
ds = ds.map(tokenize, remove_columns=ds.column_names)
quant_scheme = QuantizationScheme(
targets=["Linear"],
weights=QuantizationArgs(
num_bits=args.num_bits,
type=QuantizationType.INT,
symmetric=True,
group_size=128,
strategy=QuantizationStrategy.GROUP,
observer=args.observer,
actorder=args.actorder
),
input_activations=None,
output_activations=None,
)
recipe = [
GPTQModifier(
targets=["Linear"],
ignore=["lm_head"],
sequential_update=args.sequential_update,
dampening_frac=args.dampening_frac,
config_groups={"group_0": quant_scheme},
)
]
oneshot(
model=model,
dataset=ds,
recipe=recipe,
num_calibration_samples=args.calib_size,
)
# Save to disk compressed.
SAVE_DIR = args.quant_path
model.save_pretrained(SAVE_DIR, save_compressed=True)
tokenizer.save_pretrained(SAVE_DIR)
The model was evaluated on OpenLLM Leaderboard V1 and V2, using the following commands:
OpenLLM Leaderboard V1:
lm_eval \
--model vllm \
--model_args pretrained="neuralmagic/Mistral-Small-24B-Instruct-2501-quantized.w4a16",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,gpu_memory_utilization=0.8,enable_chunked_prefill=True,trust_remote_code=True \
--tasks openllm \
--write_out \
--batch_size auto \
--output_path output_dir \
--show_config
OpenLLM Leaderboard V2:
lm_eval \
--model vllm \
--model_args pretrained="neuralmagic/Mistral-Small-24B-Instruct-2501-quantized.w4a16",dtype=auto,add_bos_token=False,max_model_len=4096,tensor_parallel_size=1,gpu_memory_utilization=0.8,enable_chunked_prefill=True,trust_remote_code=True \
--apply_chat_template \
--fewshot_as_multiturn \
--tasks leaderboard \
--write_out \
--batch_size auto \
--output_path output_dir \
--show_config
| Metric | mistralai/Mistral-Small-24B-Instruct-2501 | neuralmagic/Mistral-Small-24B-Instruct-2501-quantized.w4a16 |
|---|---|---|
| ARC-Challenge (Acc-Norm, 25-shot) | 72.18 | 71.16 |
| GSM8K (Strict-Match, 5-shot) | 90.14 | 89.69 |
| HellaSwag (Acc-Norm, 10-shot) | 85.05 | 84.43 |
| MMLU (Acc, 5-shot) | 80.69 | 80.00 |
| TruthfulQA (MC2, 0-shot) | 65.55 | 63.92 |
| Winogrande (Acc, 5-shot) | 83.11 | 82.24 |
| Average Score | 79.45 | 78.57 |
| Recovery (%) | 100.00 | 98.9 |
| Metric | mistralai/Mistral-Small-24B-Instruct-2501 | neuralmagic/Mistral-Small-24B-Instruct-2501-quantized.w4a16 |
|---|---|---|
| IFEval (Inst-and-Prompt Level Strict Acc, 0-shot) | 73.27 | 74.37 |
| BBH (Acc-Norm, 3-shot) | 45.18 | 45.15 |
| MMLU-Pro (Acc, 5-shot) | 38.83 | 36.00 |
| Average Score | 52.42 | 51.84 |
| Recovery (%) | 100.00 | 98.89 |
| GPQA (Acc-Norm, 0-shot) | 8.29 | 6.81 |
| MUSR (Acc-Norm, 0-shot) | 7.84 | 9.46 |
Results on GPQA and MUSR are not considred for accuracy recovery calculation because the unquantized model has close to random prediction accuracy (8.29, 7.84) which doesn't provide a reliable baseline for recovery calculation.
Base model
mistralai/Mistral-Small-24B-Base-2501