|
--- |
|
tags: |
|
- int8 |
|
- vllm |
|
language: |
|
- en |
|
- de |
|
- fr |
|
- it |
|
- pt |
|
- hi |
|
- es |
|
- th |
|
pipeline_tag: text-generation |
|
license: llama3.1 |
|
base_model: meta-llama/Meta-Llama-3.1-405B-Instruct |
|
--- |
|
|
|
# Meta-Llama-3.1-405B-Instruct-quantized.w8a16 |
|
|
|
## Model Overview |
|
- **Model Architecture:** Meta-Llama-3 |
|
- **Input:** Text |
|
- **Output:** Text |
|
- **Model Optimizations:** |
|
- **Weight quantization:** INT8 |
|
- **Intended Use Cases:** Intended for commercial and research use multiple languages. Similarly to [Meta-Llama-3.1-405B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-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:** 8/19/2024 |
|
- **Version:** 1.0 |
|
- **License(s):** Llama3.1 |
|
- **Model Developers:** Neural Magic |
|
|
|
Quantized version of [Meta-Llama-3.1-405B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct). |
|
It achieves scores within 1% of the scores of the unquantized model for MMLU, ARC-Challenge, GSM-8k, Hellaswag, Winogrande and TruthfulQA. |
|
|
|
### Model Optimizations |
|
|
|
This model was obtained by quantizing the weights of [Meta-Llama-3.1-405B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-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 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. |
|
The [GPTQ](https://arxiv.org/abs/2210.17323) algorithm is applied for quantization, as implemented in the [llm-compressor](https://github.com/vllm-project/llm-compressor) library. |
|
GPTQ used a 1% damping factor and 512 sequences sequences taken from Neural Magic's [LLM compression calibration dataset](https://huggingface.co/datasets/neuralmagic/LLM_compression_calibration). |
|
|
|
|
|
## Deployment |
|
|
|
This model can be deployed efficiently using the [vLLM](https://docs.vllm.ai/en/latest/) backend, as shown in the example below. |
|
|
|
```python |
|
from vllm import LLM, SamplingParams |
|
from transformers import AutoTokenizer |
|
|
|
model_id = "neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w8a16" |
|
number_gpus = 8 |
|
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](https://docs.vllm.ai/en/latest/) for more details. |
|
|
|
|
|
## Creation |
|
|
|
This model was created by using the [llm-compressor](https://github.com/vllm-project/llm-compressor) library as presented in the code snipet below (using 8 A100 80GB GPUs). |
|
|
|
```python |
|
from transformers import AutoTokenizer |
|
from datasets import load_dataset |
|
from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot |
|
from llmcompressor.modifiers.quantization import GPTQModifier |
|
from llmcompressor.transformers.compression.helpers import custom_offload_device_map |
|
|
|
model_id = "meta-llama/Meta-Llama-3.1-405B-Instruct" |
|
|
|
num_samples = 512 |
|
max_seq_len = 4096 |
|
num_gpus = 8 |
|
max_memory_per_gpu = "20GB" |
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_id) |
|
|
|
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.shuffle().select(range(num_samples)) |
|
ds = ds.map(preprocess_fn) |
|
|
|
recipe = GPTQModifier( |
|
sequential=True |
|
targets="Linear", |
|
scheme="W8A16", |
|
ignore=["lm_head"], |
|
dampening_frac=0.01, |
|
) |
|
|
|
device_map = custom_offload_device_map( |
|
model_id, |
|
max_memory_per_gpu=max_memory_per_gpu, |
|
num_gpus=num_gpus, |
|
torch_dtype="auto", |
|
) |
|
|
|
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-3.1-405B-Instruct-quantized.w8a16") |
|
``` |
|
|
|
|
|
## Evaluation |
|
|
|
The model was evaluated on MMLU, ARC-Challenge, GSM-8K, Hellaswag, Winogrande and TruthfulQA. |
|
Evaluation was conducted using the Neural Magic fork of [lm-evaluation-harness](https://github.com/neuralmagic/lm-evaluation-harness/tree/llama_3.1_instruct) (branch llama_3.1_instruct) and the [vLLM](https://docs.vllm.ai/en/stable/) engine. |
|
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](https://huggingface.co/datasets/meta-llama/Meta-Llama-3.1-405B-Instruct-evals). |
|
|
|
### Accuracy |
|
|
|
#### Open LLM Leaderboard evaluation scores |
|
<table> |
|
<tr> |
|
<td><strong>Benchmark</strong> |
|
</td> |
|
<td><strong>Meta-Llama-3.1-405B-Instruct </strong> |
|
</td> |
|
<td><strong>Meta-Llama-3.1-405B-Instruct-quantized.w8a16 (this model)</strong> |
|
</td> |
|
<td><strong>Recovery</strong> |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>MMLU (5-shot) |
|
</td> |
|
<td>87.41 |
|
</td> |
|
<td>87.47 |
|
</td> |
|
<td>100.1% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>MMLU (CoT, 0-shot) |
|
</td> |
|
<td>88.26 |
|
</td> |
|
<td>88.23 |
|
</td> |
|
<td>100.0% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>ARC Challenge (0-shot) |
|
</td> |
|
<td>94.97 |
|
</td> |
|
<td>94.88 |
|
</td> |
|
<td>99.9% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>GSM-8K (CoT, 8-shot, strict-match) |
|
</td> |
|
<td>96.44 |
|
</td> |
|
<td>96.13 |
|
</td> |
|
<td>99.7% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>Hellaswag (10-shot) |
|
</td> |
|
<td>88.33 |
|
</td> |
|
<td>88.50 |
|
</td> |
|
<td>100.2% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>Winogrande (5-shot) |
|
</td> |
|
<td>87.21 |
|
</td> |
|
<td>87.61 |
|
</td> |
|
<td>100.5% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>TruthfulQA (0-shot, mc2) |
|
</td> |
|
<td>64.64 |
|
</td> |
|
<td>65.42 |
|
</td> |
|
<td>101.2% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td><strong>Average</strong> |
|
</td> |
|
<td><strong>86.75</strong> |
|
</td> |
|
<td><strong>86.89</strong> |
|
</td> |
|
<td><strong>100.2%</strong> |
|
</td> |
|
</tr> |
|
</table> |
|
|
|
### Reproduction |
|
|
|
The results were obtained using the following commands: |
|
|
|
#### MMLU |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=3850,max_gen_toks=10,enable_chunked_prefill=True,tensor_parallel_size=8 \ |
|
--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-3.1-405B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=4064,max_gen_toks=1024,enable_chunked_prefill=True,tensor_parallel_size=8 \ |
|
--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-3.1-405B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=3940,max_gen_toks=100,enable_chunked_prefill=True,tensor_parallel_size=8 \ |
|
--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-3.1-405B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=4096,max_gen_toks=1024,enable_chunked_prefill=True,tensor_parallel_size=8 \ |
|
--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-3.1-405B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=4096,enable_chunked_prefill=True,tensor_parallel_size=8 \ |
|
--tasks hellaswag \ |
|
--num_fewshot 10 \ |
|
--batch_size auto |
|
``` |
|
|
|
#### Winogrande |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=4096,enable_chunked_prefill=True,tensor_parallel_size=8 \ |
|
--tasks winogrande \ |
|
--num_fewshot 5 \ |
|
--batch_size auto |
|
``` |
|
|
|
#### TruthfulQA |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w8a16",dtype=auto,add_bos_token=True,max_model_len=4096,enable_chunked_prefill=True,tensor_parallel_size=8 \ |
|
--tasks truthfulqa \ |
|
--num_fewshot 0 \ |
|
--batch_size auto |
|
``` |
