--- license: llama3.2 language: - en - de - fr - it - pt - hi - es - th pipeline_tag: text-generation tags: - llama - llama-3 - neuralmagic - llmcompressor base_model: meta-llama/Llama-3.2-3B-Instruct --- # Llama-3.2-3B-Instruct-FP8 ## Model Overview - **Model Architecture:** Llama-3 - **Input:** Text - **Output:** Text - **Model Optimizations:** - **Activation quantization:** FP8 - **Weight quantization:** FP8 - **Intended Use Cases:** Intended for commercial and research use multiple languages. Similarly to [Llama-3.2-3B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-3B-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:** 9/25/2024 - **Version:** 1.0 - **License(s):** Llama3.2 - **Model Developers:** Neural Magic Quantized version of [Llama-3.2-3B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct). It achieves scores within 0.7% 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 [Llama-3.2-3B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct) to FP8 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 FP8 and floating point representations for each output channel dimension. Activations are quantized with a symmetric per-tensor scheme, where a fixed linear scaling factor is applied between FP8 and floating point representations for the entire activation tensor. Linear scaling factors are computed via by minimizing the mean squarred error (MSE). Weights are quantized by rounding to nearest FP8 representation. The [llm-compressor](https://github.com/vllm-project/llm-compressor) library was applied to quantize the model, usin 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/Llama-3.2-3B-Instruct-FP8" 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](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. ```python from transformers import AutoTokenizer from datasets import load_dataset from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot from llmcompressor.modifiers.quantization import QuantizationModifier model_id = "meta-llama/Llama-3.2-3B-Instruct" num_samples = 512 max_seq_len = 8192 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 = QuantizationModifier( targets="Linear", scheme="FP8", ignore=["lm_head"], observer="mse", ) ] 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("Llama-3.2-3B-Instruct-FP8") ``` ## 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-3.1-8B-Instruct-evals). ### Accuracy #### Open LLM Leaderboard evaluation scores
Benchmark Llama-3.2-3B-Instruct Llama-3.2-3B-Instruct-FP8 (this model) Recovery
MMLU (5-shot) 62.98 62.61 99.4%
MMLU (CoT, 0-shot) 65.40 65.20 99.7%
ARC Challenge (0-shot) 77.13 76.62 99.3%
GSM-8K (CoT, 8-shot, strict-match) 77.94 77.86 99.9%
Hellaswag (10-shot) 73.62 73.48 99.8%
Winogrande (5-shot) 71.11 70.88 99.7%
TruthfulQA (0-shot, mc2) 51.47 51.65 100.3%
Average 68.52 68.33 99.7%
### Reproduction The results were obtained using the following commands: #### MMLU ``` lm_eval \ --model vllm \ --model_args pretrained="neuralmagic/Llama-3.2-3B-Instruct-FP8",dtype=auto,add_bos_token=True,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/Llama-3.2-3B-Instruct-FP8",dtype=auto,add_bos_token=True,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/Llama-3.2-3B-Instruct-FP8",dtype=auto,add_bos_token=True,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/Llama-3.2-3B-Instruct-FP8",dtype=auto,add_bos_token=True,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/Llama-3.2-3B-Instruct-FP8",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/Llama-3.2-3B-Instruct-FP8",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/Llama-3.2-3B-Instruct-FP8",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1 \ --tasks truthfulqa \ --num_fewshot 0 \ --batch_size auto ```