Llama-2-7b-chat-quantized.w8a16

Llama-2-7b-chat-quantized.w8a16

Model Overview

• Model Architecture: Llama-2
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: INT8
• Intended Use Cases: Intended for commercial and research use in English. Similarly to Llama-2-7b-chat, 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). Use in languages other than English.
• Release Date: 7/2/2024
• Version: 1.0
• License(s): LLama2
• Model Developers: Neural Magic

Quantized version of Llama-2-7b-chat. It achieves an average score of 53.37 on the OpenLLM benchmark (version 1), whereas the unquantized model achieves 53.41.

Model Optimizations

This model was obtained by quantizing the weights of Llama-2-7b-chat to INT8 data type. This optimization reduces the number of bits per parameter from 16 to 8, reducing the disk size and GPU memory requirements by approximately 50%.

Only the weights of the linear operators within transformers blocks are quantized. Symmetric per-channel quantization is applied, in which a linear scaling per output dimension maps the INT8 and floating point representations of the quantized weights. AutoGPTQ is used for quantization with 1% damping factor and 256 sequences of 4,096 random tokens.

Deployment

Use with vLLM

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 = "neuralmagic/Llama-2-7b-chat-quantized.w8a16"
number_gpus = 1

sampling_params = SamplingParams(temperature=0.6, top_p=0.9, max_tokens=256)

tokenizer = AutoTokenizer.from_pretrained(model_id, tensor_parallel_size=number_gpus)

messages = [
    {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
    {"role": "user", "content": "Who are you?"},
]

prompts = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)

llm = LLM(model=model_id)

outputs = llm.generate(prompts, sampling_params)

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

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

Use with transformers

This model is supported by Transformers leveraging the integration with the AutoGPTQ data format. The following example contemplates how the model can be used using the generate() function.

from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "neuralmagic/Llama-2-7b-chat-quantized.w8a16"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
)

messages = [
    {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
    {"role": "user", "content": "Who are you?"},
]

input_ids = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt"
).to(model.device)

terminators = [
    tokenizer.eos_token_id,
    tokenizer.convert_tokens_to_ids("<|eot_id|>")
]

outputs = model.generate(
    input_ids,
    max_new_tokens=256,
    eos_token_id=terminators,
    do_sample=True,
    temperature=0.6,
    top_p=0.9,
)
response = outputs[0][input_ids.shape[-1]:]
print(tokenizer.decode(response, skip_special_tokens=True))

Creation

This model was created by applying the AutoGPTQ library as presented in the code snipet below. Although AutoGPTQ was used for this particular model, Neural Magic is transitioning to using llm-compressor which supports several quantization schemes and models not supported by AutoGPTQ.

from transformers import AutoTokenizer
from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
import random

model_id = "meta-llama/Llama-2-7b-chat-hf"

num_samples = 256
max_seq_len = 4096

tokenizer = AutoTokenizer.from_pretrained(model_id)

max_token_id = len(tokenizer.get_vocab()) - 1
examples = []
for _ in range(num_samples):
  examples.append(
  {
    "input_ids": [random.randint(0, max_token_id) for _ in range(max_seq_len)],
    "attention_mask": max_seq_len*[1],
  }
)

quantize_config = BaseQuantizeConfig(
  bits=8,
  group_size=-1,
  desc_act=False,
  model_file_base_name="model",
  damp_percent=0.01,
)

model = AutoGPTQForCausalLM.from_pretrained(
  model_id,
  quantize_config,
  device_map="auto",
)

model.quantize(examples)
model.save_pretrained("Llama-2-7b-chat-quantized.w8a16")

Evaluation

The model was evaluated on the OpenLLM leaderboard tasks (version 1) with the lm-evaluation-harness (commit 383bbd54bc621086e05aa1b030d8d4d5635b25e6) and the vLLM engine, using the following command:

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Llama-2-7b-chat-quantized.w8a16",dtype=auto,gpu_memory_utilization=0.4,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1 \
  --tasks openllm \
  --batch_size auto

Accuracy

Open LLM Leaderboard evaluation scores

<table> <tr> <td><strong>Benchmark</strong> </td> <td><strong>Llama-2-7b-chat </strong> </td> <td><strong>Llama-2-7b-chat-quantized.w8a16(this model)</strong> </td> <td><strong>Recovery</strong> </td> </tr> <tr> <td>MMLU (5-shot) </td> <td>47.34 </td> <td>47.32 </td> <td>100.0% </td> </tr> <tr> <td>ARC Challenge (25-shot) </td> <td>53.24 </td> <td>53.50 </td> <td>100.5% </td> </tr> <tr> <td>GSM-8K (5-shot, strict-match) </td> <td>23.20 </td> <td>22.82 </td> <td>98.4% </td> </tr> <tr> <td>Hellaswag (10-shot) </td> <td>78.65 </td> <td>78.53 </td> <td>99.8% </td> </tr> <tr> <td>Winogrande (5-shot) </td> <td>72.45 </td> <td>72.45 </td> <td>100.0% </td> </tr> <tr> <td>TruthfulQA (0-shot) </td> <td>45.58 </td> <td>45.61 </td> <td>100.1% </td> </tr> <tr> <td><strong>Average</strong> </td> <td><strong>53.41</strong> </td> <td><strong>53.37</strong> </td> <td><strong>99.9%</strong> </td> </tr> </table>