Meta-Llama-3.1-8B-Instruct-quantized.w8a8

Model Overview

• Model Architecture: Meta-Llama-3
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Activation quantization: INT8
  • Weight quantization: INT8
• Intended Use Cases: Intended for commercial and research use multiple languages. Similarly to Meta-Llama-3.1-8B-Instruct, 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: 7/11/2024
• Version: 1.0
• License(s): Llama3.1
• Model Developers: Neural Magic

This model is a quantized version of Meta-Llama-3.1-8B-Instruct. It was evaluated on a several tasks to assess the its quality in comparison to the unquatized model, including multiple-choice, math reasoning, and open-ended text generation. Meta-Llama-3.1-8B-Instruct-quantized.w8a8 achieves 105.4% recovery for the Arena-Hard evaluation, 100.3% for OpenLLM v1 (using Meta's prompting when available), 101.5% for OpenLLM v2, 99.7% for HumanEval pass@1, and 98.8% for HumanEval+ pass@1.

Model Optimizations

This model was obtained by quantizing the weights of Meta-Llama-3.1-8B-Instruct to INT8 data type. This optimization reduces the number of bits used to represent weights and activations from 16 to 8, reducing GPU memory requirements (by approximately 50%) and increasing matrix-multiply compute throughput (by approximately 2x). Weight quantization also reduces disk size requirements by approximately 50%.

Only weights and activations of the linear operators within transformers blocks are quantized. Weights are quantized with a symmetric static per-channel scheme, where a fixed linear scaling factor is applied between INT8 and floating point representations for each output channel dimension. Activations are quantized with a symmetric dynamic per-token scheme, computing a linear scaling factor at runtime for each token between INT8 and floating point representations. The GPTQ algorithm is applied for quantization, as implemented in the llm-compressor library. GPTQ used a 1% damping factor and 256 sequences of 8,192 random tokens.

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 = "neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8"
number_gpus = 1
max_model_len = 8192

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

tokenizer = AutoTokenizer.from_pretrained(model_id)

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, tensor_parallel_size=number_gpus, max_model_len=max_model_len)

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.

Creation

This model was created by using the llm-compressor library as presented in the code snipet below.

from transformers import AutoTokenizer
from datasets import Dataset
from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot
from llmcompressor.modifiers.quantization import GPTQModifier
import random

model_id = "meta-llama/Meta-Llama-3.1-8B-Instruct"

num_samples = 256
max_seq_len = 8192

tokenizer = AutoTokenizer.from_pretrained(model_id)

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

recipe = GPTQModifier(
  targets="Linear",
  scheme="W8A8",
  ignore=["lm_head"],
  dampening_frac=0.01,
)

model = SparseAutoModelForCausalLM.from_pretrained(
  model_id,
  device_map="auto",
)

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

model.save_pretrained("Meta-Llama-3.1-8B-Instruct-quantized.w8a8")

Evaluation

This model was evaluated on the well-known Arena-Hard, OpenLLM v1, OpenLLM v2, HumanEval, and HumanEval+ benchmarks. In all cases, model outputs were generated with the vLLM engine.

Arena-Hard evaluations were conducted using the Arena-Hard-Auto repository. The model generated a single answer for each prompt form Arena-Hard, and each answer was judged twice by GPT-4. We report below the scores obtained in each judgement and the average.

OpenLLM v1 and v2 evaluations were conducted using Neural Magic's fork of lm-evaluation-harness (branch llama_3.1_instruct). This version of the lm-evaluation-harness includes versions of MMLU, ARC-Challenge and GSM-8K that match the prompting style of Meta-Llama-3.1-Instruct-evals and a few fixes to OpenLLM v2 tasks.

HumanEval and HumanEval+ evaluations were conducted using Neural Magic's fork of the EvalPlus repository.

Detailed model outputs are available as HuggingFace datasets for Arena-Hard, OpenLLM v2, and HumanEval.

Note: Results have been updated after Meta modified the chat template.

Accuracy

Open LLM Leaderboard evaluation scores

Benchmark	Meta-Llama-3.1-8B-Instruct	Meta-Llama-3.1-8B-Instruct-quantized.w8a8 (this model)	Recovery
Arena Hard	25.8 (25.1 / 26.5)	27.2 (27.6 / 26.7)	105.4%
OpenLLM v1
MMLU (5-shot)	68.3	67.8	99.3%
MMLU (CoT, 0-shot)	72.8	72.2	99.1%
ARC Challenge (0-shot)	81.4	81.7	100.3%
GSM-8K (CoT, 8-shot, strict-match)	82.8	84.8	102.5%
Hellaswag (10-shot)	80.5	80.3	99.8%
Winogrande (5-shot)	78.1	78.5	100.5%
TruthfulQA (0-shot, mc2)	54.5	54.7	100.3%
Average	74.1	74.3	100.3%
OpenLLM v2
MMLU-Pro (5-shot)	30.8	30.9	100.3%
IFEval (0-shot)	77.9	78.0	100.1%
BBH (3-shot)	30.1	31.0	102.9%
Math-|v|-5 (4-shot)	15.7	15.5	98.9%
GPQA (0-shot)	3.7	5.4	146.2%
MuSR (0-shot)	7.6	7.6	100.0%
Average	27.6	28.0	101.5%
Coding
HumanEval pass@1	67.3	67.1	99.7%
HumanEval+ pass@1	60.7	60.0	98.8%

Reproduction

The results were obtained using the following commands:

MMLU

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",dtype=auto,max_model_len=3850,max_gen_toks=10,tensor_parallel_size=1 \
  --tasks mmlu_llama_3.1_instruct \
  --fewshot_as_multiturn \
  --apply_chat_template \
  --num_fewshot 5 \
  --batch_size auto

MMLU-CoT

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",dtype=auto,max_model_len=4064,max_gen_toks=1024,tensor_parallel_size=1 \
  --tasks mmlu_cot_0shot_llama_3.1_instruct \
  --apply_chat_template \
  --num_fewshot 0 \
  --batch_size auto

ARC-Challenge

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",dtype=auto,max_model_len=3940,max_gen_toks=100,tensor_parallel_size=1 \
  --tasks arc_challenge_llama_3.1_instruct \
  --apply_chat_template \
  --num_fewshot 0 \
  --batch_size auto

GSM-8K

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",dtype=auto,max_model_len=4096,max_gen_toks=1024,tensor_parallel_size=1 \
  --tasks gsm8k_cot_llama_3.1_instruct \
  --fewshot_as_multiturn \
  --apply_chat_template \
  --num_fewshot 8 \
  --batch_size auto

Hellaswag

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1 \
  --tasks hellaswag \
  --num_fewshot 10 \
  --batch_size auto

Winogrande

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1 \
  --tasks winogrande \
  --num_fewshot 5 \
  --batch_size auto

TruthfulQA

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1 \
  --tasks truthfulqa \
  --num_fewshot 0 \
  --batch_size auto

OpenLLM v2

lm_eval \
  --model vllm \
  --model_args pretrained="neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",dtype=auto,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --tasks leaderboard \
  --batch_size auto

HumanEval and HumanEval+

Generation

python3 codegen/generate.py \
  --model neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w8a8 \
  --bs 16 \
  --temperature 0.2 \
  --n_samples 50 \
  --root "." \
  --dataset humaneval

Sanitization

python3 evalplus/sanitize.py \
  humaneval/neuralmagic--Meta-Llama-3.1-8B-Instruct-quantized.w8a8_vllm_temp_0.2

Evaluation

evalplus.evaluate \
  --dataset humaneval \
  --samples humaneval/neuralmagic--Meta-Llama-3.1-8B-Instruct-quantized.w8a8_vllm_temp_0.2-sanitized