Qwen2.5-7B-quantized.w4a16

Qwen2.5-7B-quantized.w4a16

Model Overview

• Model Architecture: Qwen2
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: INT4
• Intended Use Cases: Intended for commercial and research use multiple languages. Similarly to Qwen2.5-7B, this models is intended for assistant-like chat.
• Out-of-scope: Use in any manner that violates applicable laws or regulations (including trade compliance laws).
• Release Date: 10/18/2024
• Version: 1.0
• License(s): apache-2.0
• Model Developers: Neural Magic

Model Optimizations

This model was obtained by quantizing the weights of Qwen2.5-7B to INT4 data type. 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. Weights are quantized using a symmetric per-group scheme, with group size 128. The GPTQ algorithm is applied for quantization, as implemented in the llm-compressor library.

Deployment

This model can be deployed efficiently using the vLLM backend, as shown in the example below.

from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

model_id = "RedHatAI/Qwen2.5-7B-quantized.w4a16"
number_gpus = 1
max_model_len = 8192

sampling_params = SamplingParams(temperature=0.7, top_p=0.8, max_tokens=256)

tokenizer = AutoTokenizer.from_pretrained(model_id)

prompt = "Give me a short introduction to large language model."

llm = LLM(model=model_id, tensor_parallel_size=number_gpus, max_model_len=max_model_len)

outputs = llm.generate(prompt, sampling_params)

generated_text = outputs[0].outputs[0].text
print(generated_text)

vLLM aslo supports OpenAI-compatible serving. See the documentation for more details.

Creation

<details> <summary>Creation details</summary> This model was created with [llm-compressor](https://github.com/vllm-project/llm-compressor) by running the code snippet below.

from transformers import AutoModelForCausalLM, AutoTokenizer
from llmcompressor.modifiers.quantization import GPTQModifier
from llmcompressor.transformers import oneshot
from datasets import load_dataset

# Load model
model_stub = "Qwen/Qwen2.5-7B"
model_name = model_stub.split("/")[-1]

num_samples = 3072
max_seq_len = 8192

tokenizer = AutoTokenizer.from_pretrained(model_stub)

model = AutoModelForCausalLM.from_pretrained(
    model_stub,
    device_map="auto",
    torch_dtype="auto",
)

def preprocess_fn(example):
  return {"text": tokenizer.apply_chat_template(example["messages"], add_generation_prompt=False, tokenize=False)}

ds = load_dataset("neuralmagic/LLM_compression_calibration", split="train")
ds = ds.map(preprocess_fn)

# Configure the quantization algorithm and scheme
recipe = GPTQModifier(
    targets="Linear",
    scheme="W4A16",
    ignore=["lm_head"],
    sequential_targets=["Qwen2DecoderLayer"],
    dampening_frac=0.1,
)

# Apply quantization
oneshot(
    model=model,
    dataset=ds, 
    recipe=recipe,
    max_seq_length=max_seq_len,
    num_calibration_samples=num_samples,
)

# Save to disk in compressed-tensors format
save_path = model_name + "-quantized.w4a16"
model.save_pretrained(save_path)
tokenizer.save_pretrained(save_path)
print(f"Model and tokenizer saved to: {save_path}")

</details>

Evaluation

The model was evaluated on the OpenLLM leaderboard tasks (version 1) with the lm-evaluation-harness (commit 387Bbd54bc621086e05aa1b030d8d4d5635b25e6) and the vLLM engine, using the following command:

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Qwen2.5-7B-quantized.w4a16",dtype=auto,gpu_memory_utilization=0.9,add_bos_token=True,max_model_len=4096,enable_chunk_prefill=True,tensor_parallel_size=1 \
  --tasks openllm \
  --batch_size auto

Accuracy

Open LLM Leaderboard evaluation scores

<table> <tr> <th>Benchmark </th> <th>Qwen2.5-7B </th> <th>Qwen2.5-7B-quantized.w4a16<br>(this model) </th> <th>Recovery </th> </tr> <tr> <td>MMLU (5-shot) </td> <td>74.15 </td> <td>73.47 </td> <td>99.1% </td> </tr> <tr> <td>ARC Challenge (25-shot) </td> <td>59.39 </td> <td>58.70 </td> <td>98.9% </td> </tr> <tr> <td>GSM-8K (5-shot, strict-match) </td> <td>79.76 </td> <td>79.08 </td> <td>99.1% </td> </tr> <tr> <td>Hellaswag (10-shot) </td> <td>80.17 </td> <td>79.39 </td> <td>99.0% </td> </tr> <tr> <td>Winogrande (5-shot) </td> <td>75.69 </td> <td>76.01 </td> <td>100.4% </td> </tr> <tr> <td>TruthfulQA (0-shot, mc2) </td> <td>56.38 </td> <td>55.48 </td> <td>98.4% </td> </tr> <tr> <td><strong>Average</strong> </td> <td><strong>70.92</strong> </td> <td><strong>70.35</strong> </td> <td><strong>99.2%</strong> </td> </tr> </table>