|
--- |
|
tags: |
|
- fp8 |
|
- vllm |
|
language: |
|
- en |
|
- de |
|
- fr |
|
- it |
|
- pt |
|
- hi |
|
- es |
|
- th |
|
pipeline_tag: text-generation |
|
license: llama3.1 |
|
base_model: meta-llama/Meta-Llama-3.1-70B-Instruct |
|
--- |
|
|
|
# Meta-Llama-3.1-70B-Instruct-FP8-dynamic |
|
|
|
## Model Overview |
|
- **Model Architecture:** Meta-Llama-3.1 |
|
- **Input:** Text |
|
- **Output:** Text |
|
- **Model Optimizations:** |
|
- **Weight quantization:** FP8 |
|
- **Activation quantization:** FP8 |
|
- **Intended Use Cases:** Intended for commercial and research use in multiple languages. Similarly to [Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/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). Use in languages other than English. |
|
- **Release Date:** 7/23/2024 |
|
- **Version:** 1.0 |
|
- **License(s):** [llama3.1](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B/blob/main/LICENSE) |
|
- **Model Developers:** Neural Magic |
|
|
|
This model is a quantized version of [Meta-Llama-3.1-70B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-70B-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-3.1-70B-Instruct-FP8-dynamic achieves 101.6% recovery for the Arena-Hard evaluation, 99.7% for OpenLLM v1 (using Meta's prompting when available), 100.0% for OpenLLM v2, 100.4% for HumanEval pass@1, and 100.3% for HumanEval+ pass@1. |
|
|
|
### Model Optimizations |
|
|
|
This model was obtained by quantizing the weights and activations of [Meta-Llama-3.1-70B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-70B-Instruct) to FP8 data type, ready for inference with vLLM built from source. |
|
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 and activations 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 FP8 representations of the quantized weights and activations. Activations are also quantized on a per-token dynamic basis. |
|
[LLM Compressor](https://github.com/vllm-project/llm-compressor) is used for quantization. |
|
|
|
## 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 vllm import LLM, SamplingParams |
|
from transformers import AutoTokenizer |
|
|
|
model_id = "neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8-dynamic" |
|
number_gpus = 2 |
|
|
|
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) |
|
|
|
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 applying [LLM Compressor with calibration samples from UltraChat](https://github.com/vllm-project/llm-compressor/blob/sa/big_model_support/examples/big_model_offloading/big_model_w8a8_calibrate.py), as presented in the code snipet below. |
|
|
|
```python |
|
import torch |
|
|
|
from transformers import AutoTokenizer |
|
|
|
from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot |
|
from llmcompressor.transformers.compression.helpers import ( # noqa |
|
calculate_offload_device_map, |
|
custom_offload_device_map, |
|
) |
|
|
|
recipe = """ |
|
quant_stage: |
|
quant_modifiers: |
|
QuantizationModifier: |
|
ignore: ["lm_head"] |
|
config_groups: |
|
group_0: |
|
weights: |
|
num_bits: 8 |
|
type: float |
|
strategy: channel |
|
dynamic: false |
|
symmetric: true |
|
input_activations: |
|
num_bits: 8 |
|
type: float |
|
strategy: token |
|
dynamic: true |
|
symmetric: true |
|
targets: ["Linear"] |
|
""" |
|
|
|
model_stub = "meta-llama/Meta-Llama-3.1-70B-Instruct" |
|
model_name = model_stub.split("/")[-1] |
|
|
|
device_map = calculate_offload_device_map( |
|
model_stub, reserve_for_hessians=False, num_gpus=2, torch_dtype="auto" |
|
) |
|
|
|
model = SparseAutoModelForCausalLM.from_pretrained( |
|
model_stub, torch_dtype="auto", device_map=device_map |
|
) |
|
|
|
output_dir = f"./{model_name}-FP8-dynamic" |
|
|
|
oneshot( |
|
model=model, |
|
recipe=recipe, |
|
output_dir=output_dir, |
|
save_compressed=True, |
|
tokenizer=AutoTokenizer.from_pretrained(model_stub), |
|
) |
|
``` |
|
|
|
## 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](https://docs.vllm.ai/en/stable/) engine. |
|
|
|
Arena-Hard evaluations were conducted using the [Arena-Hard-Auto](https://github.com/lmarena/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](https://github.com/neuralmagic/lm-evaluation-harness/tree/llama_3.1_instruct) (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](https://huggingface.co/datasets/meta-llama/Meta-Llama-3.1-70B-Instruct-evals) and a few fixes to OpenLLM v2 tasks. |
|
|
|
HumanEval and HumanEval+ evaluations were conducted using Neural Magic's fork of the [EvalPlus](https://github.com/neuralmagic/evalplus) repository. |
|
|
|
Detailed model outputs are available as HuggingFace datasets for [Arena-Hard](https://huggingface.co/datasets/neuralmagic/quantized-llama-3.1-arena-hard-evals), [OpenLLM v2](https://huggingface.co/datasets/neuralmagic/quantized-llama-3.1-leaderboard-v2-evals), and [HumanEval](https://huggingface.co/datasets/neuralmagic/quantized-llama-3.1-humaneval-evals). |
|
|
|
### Accuracy |
|
|
|
<table> |
|
<tr> |
|
<td><strong>Benchmark</strong> |
|
</td> |
|
<td><strong>Meta-Llama-3.1-70B-Instruct </strong> |
|
</td> |
|
<td><strong>Meta-Llama-3.1-70B-Instruct-FP8-dynamic (this model)</strong> |
|
</td> |
|
<td><strong>Recovery</strong> |
|
</td> |
|
</tr> |
|
<tr> |
|
<td><strong>Arena Hard</strong> |
|
</td> |
|
<td>57.0 (55.8 / 58.2) |
|
</td> |
|
<td>57.9 (58.1 / 57.7) |
|
</td> |
|
<td>101.6% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td><strong>OpenLLM v1</strong> |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>MMLU (5-shot) |
|
</td> |
|
<td>83.8 |
|
</td> |
|
<td>83.8 |
|
</td> |
|
<td>99.9% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>MMLU-cot (0-shot) |
|
</td> |
|
<td>86.0 |
|
</td> |
|
<td>85.5 |
|
</td> |
|
<td>99.4% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>ARC Challenge (0-shot) |
|
</td> |
|
<td>93.3 |
|
</td> |
|
<td>93.5 |
|
</td> |
|
<td>100.2% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>GSM-8K-cot (8-shot, strict-match) |
|
</td> |
|
<td>94.9 |
|
</td> |
|
<td>94.5 |
|
</td> |
|
<td>99.6% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>Hellaswag (10-shot) |
|
</td> |
|
<td>86.8 |
|
</td> |
|
<td>86.6 |
|
</td> |
|
<td>99.9% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>Winogrande (5-shot) |
|
</td> |
|
<td>85.3 |
|
</td> |
|
<td>84.6 |
|
</td> |
|
<td>99.2% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>TruthfulQA (0-shot, mc2) |
|
</td> |
|
<td>60.7 |
|
</td> |
|
<td>60.6 |
|
</td> |
|
<td>99.9% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td><strong>Average</strong> |
|
</td> |
|
<td><strong>84.4</strong> |
|
</td> |
|
<td><strong>84.2</strong> |
|
</td> |
|
<td><strong>99.7%</strong> |
|
</td> |
|
</tr> |
|
<tr> |
|
<td><strong>OpenLLM v2</strong> |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>MMLU-Pro (5-shot) |
|
</td> |
|
<td>48.1 |
|
</td> |
|
<td>47.7 |
|
</td> |
|
<td>99.1% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>IFEval (0-shot) |
|
</td> |
|
<td>86.4 |
|
</td> |
|
<td>87.6 |
|
</td> |
|
<td>101.3% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>BBH (3-shot) |
|
</td> |
|
<td>55.8 |
|
</td> |
|
<td>54.9 |
|
</td> |
|
<td>98.4% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>Math-|v|-5 (4-shot) |
|
</td> |
|
<td>26.1 |
|
</td> |
|
<td>28.0 |
|
</td> |
|
<td>107.5% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>GPQA (0-shot) |
|
</td> |
|
<td>15.4 |
|
</td> |
|
<td>14.6 |
|
</td> |
|
<td>94.7% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>MuSR (0-shot) |
|
</td> |
|
<td>18.2 |
|
</td> |
|
<td>17.2 |
|
</td> |
|
<td>94.5% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td><strong>Average</strong> |
|
</td> |
|
<td><strong>41.7</strong> |
|
</td> |
|
<td><strong>41.7</strong> |
|
</td> |
|
<td><strong>100.0%</strong> |
|
</td> |
|
</tr> |
|
<tr> |
|
<td><strong>Coding</strong> |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>HumanEval pass@1 |
|
</td> |
|
<td>79.7 |
|
</td> |
|
<td>80.0 |
|
</td> |
|
<td>100.4% |
|
</td> |
|
</tr> |
|
<tr> |
|
<td>HumanEval+ pass@1 |
|
</td> |
|
<td>74.8 |
|
</td> |
|
<td>75.0 |
|
</td> |
|
<td>100.3% |
|
</td> |
|
</tr> |
|
</table> |
|
|
|
### Reproduction |
|
|
|
The results were obtained using the following commands: |
|
|
|
#### MMLU |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=2 \ |
|
--tasks mmlu \ |
|
--num_fewshot 5 \ |
|
--batch_size auto |
|
``` |
|
|
|
#### MMLU-cot |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=2 \ |
|
--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-3.1-70B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=2 \ |
|
--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-3.1-70B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=2 \ |
|
--tasks gsm8k_cot_llama_3.1_instruct \ |
|
--apply_chat_template \ |
|
--fewshot_as_multiturn \ |
|
--num_fewshot 8 \ |
|
--batch_size auto |
|
``` |
|
|
|
#### Hellaswag |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=2 \ |
|
--tasks hellaswag \ |
|
--num_fewshot 10 \ |
|
--batch_size auto |
|
``` |
|
|
|
#### Winogrande |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=2 \ |
|
--tasks winogrande \ |
|
--num_fewshot 5 \ |
|
--batch_size auto |
|
``` |
|
|
|
#### TruthfulQA |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8-dynamic",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=2 \ |
|
--tasks truthfulqa \ |
|
--num_fewshot 0 \ |
|
--batch_size auto |
|
``` |
|
|
|
#### OpenLLM v2 |
|
``` |
|
lm_eval \ |
|
--model vllm \ |
|
--model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8-dynamic",dtype=auto,max_model_len=4096,tensor_parallel_size=2,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-3.1-70B-Instruct-FP8-dynamic \ |
|
--bs 16 \ |
|
--temperature 0.2 \ |
|
--n_samples 50 \ |
|
--root "." \ |
|
--dataset humaneval \ |
|
--tp 2 |
|
``` |
|
##### Sanitization |
|
``` |
|
python3 evalplus/sanitize.py \ |
|
humaneval/neuralmagic--Meta-Llama-3.1-70B-Instruct-FP8-dynamic_vllm_temp_0.2 |
|
``` |
|
##### Evaluation |
|
``` |
|
evalplus.evaluate \ |
|
--dataset humaneval \ |
|
--samples humaneval/neuralmagic--Meta-Llama-3.1-70B-Instruct-FP8-dynamic_vllm_temp_0.2-sanitized |
|
``` |
