--- tags: - fp8 - vllm license: apache-2.0 --- # Mixtral-8x22B-Instruct-v0.1-FP8 ## Model Overview - **Model Architecture:** Mixtral-8x22B-Instruct-v0.1 - **Input:** Text - **Output:** Text - **Model Optimizations:** - **Weight quantization:** FP8 - **Activation quantization:** FP8 - **Intended Use Cases:** Intended for commercial and research use in English. Similarly to [Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-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:** 8/11/2024 - **Version:** 1.1 - **License(s):** [apache-2.0](https://huggingface.co/datasets/choosealicense/licenses/blob/main/markdown/apache-2.0.md) - **Model Developers:** Neural Magic Quantized version of [Mixtral-8x22B-Instruct-v0.1](https://huggingface.co/mistralai/Mixtral-8x22B-Instruct-v0.1) with the updated tokenizer. It achieves an average score of 79.04 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 79.93. ### Model Optimizations This model was obtained by quantizing the weights and activations of [Mixtral-8x22B-Instruct-v0.1](https://huggingface.co/mistralai/Mixtral-8x22B-Instruct-v0.1) to FP8 data type, ready for inference with vLLM >= 0.5.0. 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-tensor quantization is applied, in which a single linear scaling maps the FP8 representations of the quantized weights and activations. [AutoFP8](https://github.com/neuralmagic/AutoFP8) is used for quantization with 512 sequences of UltraChat. ## 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/Mixtral-8x22B-Instruct-v0.1-FP8" 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, 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](https://docs.vllm.ai/en/latest/) for more details. ## Creation This model was created by applying [AutoFP8 with calibration samples from ultrachat](https://github.com/neuralmagic/AutoFP8/blob/147fa4d9e1a90ef8a93f96fc7d9c33056ddc017a/example_dataset.py) with block_sparse_moe.gate layers kept at original precision, as presented in the code snipet below. Although AutoFP8 was used for this particular model, Neural Magic is transitioning to using [llm-compressor](https://github.com/vllm-project/llm-compressor) which supports several quantization schemes and models not supported by AutoFP8. ```python from datasets import load_dataset from transformers import AutoTokenizer from auto_fp8 import AutoFP8ForCausalLM, BaseQuantizeConfig pretrained_model_dir = "mistralai/Mixtral-8x22B-Instruct-v0.1" quantized_model_dir = "Mixtral-8x22B-Instruct-v0.1-FP8" tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True, model_max_length=4096) tokenizer.pad_token = tokenizer.eos_token ds = load_dataset("mgoin/ultrachat_2k", split="train_sft").select(range(512)) examples = [tokenizer.apply_chat_template(batch["messages"], tokenize=False) for batch in ds] examples = tokenizer(examples, padding=True, truncation=True, return_tensors="pt").to("cuda") quantize_config = BaseQuantizeConfig( quant_method="fp8", activation_scheme="static", ignore_patterns=["re:.*lm_head", "re:.*block_sparse_moe.gate"], ) model = AutoFP8ForCausalLM.from_pretrained( pretrained_model_dir, quantize_config=quantize_config ) model.quantize(examples) model.save_quantized(quantized_model_dir) ``` ## Evaluation The model was evaluated on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) leaderboard tasks (version 1) with the [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness/tree/383bbd54bc621086e05aa1b030d8d4d5635b25e6) (commit 383bbd54bc621086e05aa1b030d8d4d5635b25e6) and the [vLLM](https://docs.vllm.ai/en/stable/) engine, using the following command: ``` lm_eval \ --model vllm \ --model_args pretrained="neuralmagic/Mixtral-8x22B-Instruct-v0.1-FP8",tensor_parallel_size=4,dtype=auto,gpu_memory_utilization=0.8,add_bos_token=True,max_model_len=4096 \ --tasks openllm \ --batch_size auto ``` ### Accuracy #### Open LLM Leaderboard evaluation scores
Benchmark Mixtral-8x22B-Instruct-v0.1 Mixtral-8x22B-Instruct-v0.1-FP8(this model) Recovery
MMLU (5-shot) 77.71 77.03 99.12%
ARC Challenge (25-shot) 73.38 73.04 99.54%
GSM-8K (5-shot, strict-match) 84.99 83.62 98.39%
Hellaswag (10-shot) 89.24 88.22 98.86%
Winogrande (5-shot) 85.87 84.93 98.91%
TruthfulQA (0-shot) 68.41 67.37 98.48%
Average 79.93 79.04 98.88%