|
|
--- |
|
|
tags: |
|
|
- w4a16 |
|
|
- int4 |
|
|
- vllm |
|
|
license: apache-2.0 |
|
|
license_link: https://huggingface.co/Qwen/QwQ-32B-Preview/blob/main/LICENSE |
|
|
language: |
|
|
- en |
|
|
base_model: Qwen/Qwen2.5-32B-Instruct |
|
|
--- |
|
|
|
|
|
# QwQ-32B-Preview-quantized.w4a16 |
|
|
|
|
|
## Model Overview |
|
|
- **Model Architecture:** QwQ-32B-Preview |
|
|
- **Input:** Text |
|
|
- **Output:** Text |
|
|
- **Model Optimizations:** |
|
|
- **Weight quantization:** INT4 |
|
|
- **Activation quantization:** None |
|
|
- **Release Date:** 3/1/2025 |
|
|
- **Version:** 1.0 |
|
|
- **Model Developers:** Neural Magic |
|
|
|
|
|
Quantized version of [QwQ-32B-Preview](https://huggingface.co/Qwen/QwQ-32B-Preview). |
|
|
It achieves an average score of 75.87 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 77.20. |
|
|
|
|
|
### Model Optimizations |
|
|
|
|
|
This model was obtained by only quantizing the weights of [QwQ-32B-Preview](https://huggingface.co/Qwen/QwQ-32B-Preview) to INT4 data type, ready for inference with vLLM >= 0.5.2. |
|
|
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. |
|
|
|
|
|
## Deployment |
|
|
|
|
|
### Use with vLLM |
|
|
|
|
|
This model can be deployed efficiently using the [vLLM](https://docs.vllm.ai/en/latest/) backend, as shown in the example below. |
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer |
|
|
from vllm import LLM, SamplingParams |
|
|
|
|
|
max_model_len, tp_size = 4096, 1 |
|
|
model_name = "neuralmagic-ent/QwQ-32B-Preview-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](https://docs.vllm.ai/en/latest/) for more details. |
|
|
|
|
|
## Creation |
|
|
|
|
|
This model was created with [llm-compressor](https://github.com/vllm-project/llm-compressor) by running the code snippet below with the following arguments: |
|
|
|
|
|
|
|
|
```bash |
|
|
python quantize.py --model_path Qwen/QwQ-32B-Preview --quant_path "output_dir/QwQ-32B-Preview-quantized.w4a16" --calib_size 128 --dampening_frac 0.1 --observer minmax --actorder False |
|
|
``` |
|
|
|
|
|
|
|
|
```python |
|
|
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" |
|
|
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, |
|
|
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) |
|
|
``` |
|
|
|
|
|
## Evaluation |
|
|
|
|
|
The model was evaluated on OpenLLM Leaderboard [V1](https://huggingface.co/spaces/open-llm-leaderboard-old/open_llm_leaderboard) and [V2](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard#/), using the following commands: |
|
|
|
|
|
OpenLLM Leaderboard V1: |
|
|
``` |
|
|
lm_eval \ |
|
|
--model vllm \ |
|
|
--model_args pretrained="neuralmagic-ent/QwQ-32B-Preview-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-ent/QwQ-32B-Preview-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 |
|
|
|
|
|
``` |
|
|
|
|
|
### Accuracy |
|
|
|
|
|
#### OpenLLM Leaderboard V1 evaluation scores |
|
|
|
|
|
| Metric | Qwen/QwQ-32B-Preview | neuralmagic-ent/QwQ-32B-Preview-quantized.w4a16 | |
|
|
|-----------------------------------------|:---------------------------------:|:-------------------------------------------:| |
|
|
| ARC-Challenge (Acc-Norm, 25-shot) | 70.73 | 70.14 | |
|
|
| GSM8K (Strict-Match, 5-shot) | 83.09 | 79.83 | |
|
|
| HellaSwag (Acc-Norm, 10-shot) | 85.77 | 85.01 | |
|
|
| MMLU (Acc, 5-shot) | 82.67 | 81.58 | |
|
|
| TruthfulQA (MC2, 0-shot) | 60.88 | 59.57 | |
|
|
| Winogrande (Acc, 5-shot) | 80.03 | 79.08 | |
|
|
| **Average Score** | **77.20** | **75.87** | |
|
|
| **Recovery** | **100.00** | **98.28** | |
|
|
|
|
|
#### OpenLLM Leaderboard V2 evaluation scores |
|
|
|
|
|
| Metric | Qwen/QwQ-32B-Preview | neuralmagic-ent/QwQ-32B-Preview-quantized.w4a16 | |
|
|
|---------------------------------------------------------|:---------------------------------:|:-------------------------------------------:| |
|
|
| IFEval (Inst-and-Prompt Level Strict Acc, 0-shot) | 42.34 | 41.09 | |
|
|
| BBH (Acc-Norm, 3-shot) | 53.03 | 50.19 | |
|
|
| Math-Hard (Exact-Match, 4-shot) | 21.15 | 22.68 | |
|
|
| GPQA (Acc-Norm, 0-shot) | 2.97 | 3.64 | |
|
|
| MUSR (Acc-Norm, 0-shot) | 9.57 | 14.01 | |
|
|
| MMLU-Pro (Acc, 5-shot) | 52.00 | 49.87 | |
|
|
| **Average Score** | **30.18** | **30.25** | |
|
|
| **Recovery** | **100.00** | **100.23** | |
|
|
|
|
|
|
