--- library_name: transformers license: apache-2.0 language: - en pipeline_tag: text-generation tags: - int8 - vllm base_model: HuggingFaceTB/SmolLM-1.7B-Instruct --- # SmolLM-1.7B-Instruct-quantized.w8a8 ## Model Overview - **Model Architecture:** Llama - **Input:** Text - **Output:** Text - **Model Optimizations:** - **Activation quantization:** INT8 - **Weight quantization:** INT8 - **Intended Use Cases:** Intended for commercial and research use in English. Similarly to [SmolLM-1.7B-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM-1.7B-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:** 8/23/2024 - **Version:** 1.0 - **License(s):** [Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0) - **Model Developers:** Neural Magic Quantized version of [SmolLM-1.7B-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM-1.7B-Instruct). It achieves an average score of 41.23 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 41.76. ### Model Optimizations This model was obtained by quantizing the weights of [SmolLM-1.7B-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM-1.7B-Instruct) to INT8 data type. 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 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](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 1,024 sequences sequences taken from Neural Magic's [LLM compression calibration dataset](https://huggingface.co/datasets/neuralmagic/LLM_compression_calibration). ## 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/SmolLM-1.7B-Instruct-quantized.w8a8" sampling_params = SamplingParams(temperature=0.6, top_p=0.92, max_tokens=100) tokenizer = AutoTokenizer.from_pretrained(model_id) messages = [ {"role": "user", "content": "List the steps to bake a chocolate cake from scratch."}, ] prompts = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) llm = LLM(model=model_id) outputs = llm.generate(prompts, sampling_params) generated_text = outputs[0].outputs[0].text print(generated_text) ``` vLLM also 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 Dataset from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot from llmcompressor.modifiers.quantization import GPTQModifier import random model_id = "HuggingFaceTB/SmolLM-1.7B-Instruct" num_samples = 1024 max_seq_len = 2048 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( 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("SmolLM-1.7B-Instruct-quantized.w8a8") ``` ## 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/SmolLM-1.7B-Instruct-quantized.w8a8",dtype=auto,gpu_memory_utilization=0.4,add_bos_token=True,max_model_len=4096 \ --tasks openllm \ --batch_size auto ``` ### Accuracy #### Open LLM Leaderboard evaluation scores
Benchmark SmolLM-1.7B-Instruct-quantized SmolLM-1.7B-Instruct-quantized.w8a8 (this model) Recovery
MMLU (5-shot) 28.10 27.54 98.0%
ARC Challenge (25-shot) 49.06 48.98 99.8%
GSM-8K (5-shot, strict-match) 4.93 3.87 78.5%
Hellaswag (10-shot) 66.96 66.25 98.9%
Winogrande (5-shot) 61.01 60.54 99.2%
TruthfulQA (0-shot) 40.48 40.21 99.3%
Average 41.76 41.23 98.7%