Update README.md

README.md

@@ -7,22 +7,189 @@ tags:
 # Mixtral-8x22B-Instruct-v0.1-FP8
 
 ## Model Overview
-Mixtral-8x22B-Instruct-v0.1 quantized to FP8 weights and activations using per-tensor quantization, ready for inference with vLLM >= 0.5.0.
+- **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-3-7B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-7B-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:** 6/8/2024
+- **Version:** 1.0
+- **Model Developers:** Neural Magic
 
-## Usage and Creation
-Produced using [AutoFP8 with calibration samples from ultrachat](https://github.com/neuralmagic/AutoFP8/blob/147fa4d9e1a90ef8a93f96fc7d9c33056ddc017a/example_dataset.py).
+Quantized version of [Mixtral-8x22B-Instruct-v0.1](mistralai/Mixtral-8x22B-Instruct-v0.1).
+It achieves an average score of 78.47 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 79.15.
+
+### Model Optimizations
+
+This model was obtained by quantizing the weights and activations of [Mixtral-8x22B-Instruct-v0.1](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-channel quantization is applied, in which a linear scaling per output dimension 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
 
-### Open LLM Leaderboard evaluation scores
-|                      | Mixtral-8x22B-Instruct-v0.1 | Mixtral-8x22B-Instruct-v0.1-FP8<br>(this model) |
-| :------------------: | :----------------------: | :------------------------------------------------: |
-| arc-c<br>25-shot (acc_norm)     | 72.70                    | 72.53                                              |
-| hellaswag<br>10-shot (acc_norm) | 89.08                    | 88.10                                              |
-| mmlu<br>5-shot                  | 77.77                    | 76.08                                              |
-| truthfulqa<br>0-shot (acc)      | 68.14                    | 66.32                                              |
-| winogrande<br>5-shot (acc)      | 85.16                    | 84.37                                              |
-| gsm8k<br>5-shot (strict-match)  | 82.03                    | 83.40                                              |
-| **Average<br>Accuracy**         | **79.15**                |              **78.47**                             |
-| **Recovery**                    | **100%**                 |              **99.14%**                            |
+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",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
+<table>
+  <tr>
+   <td><strong>Benchmark</strong>
+   </td>
+   <td><strong>Mixtral-8x22B-Instruct-v0.1</strong>
+   </td>
+   <td><strong>Mixtral-8x22B-Instruct-v0.1-FP8(this model)</strong>
+   </td>
+   <td><strong>Recovery</strong>
+   </td>
+  </tr>
+  <tr>
+   <td>MMLU (5-shot)
+   </td>
+   <td>77.77
+   </td>
+   <td>76.08
+   </td>
+   <td>97.82%
+   </td>
+  </tr>
+  <tr>
+   <td>ARC Challenge (25-shot)
+   </td>
+   <td>72.70
+   </td>
+   <td>72.53
+   </td>
+   <td>99.76%
+   </td>
+  </tr>
+  <tr>
+   <td>GSM-8K (5-shot, strict-match)
+   </td>
+   <td>82.03
+   </td>
+   <td>83.40
+   </td>
+   <td>101.6%
+   </td>
+  </tr>
+  <tr>
+   <td>Hellaswag (10-shot)
+   </td>
+   <td>89.08
+   </td>
+   <td>88.10
+   </td>
+   <td>98.89%
+   </td>
+  </tr>
+  <tr>
+   <td>Winogrande (5-shot)
+   </td>
+   <td>85.16
+   </td>
+   <td>84.37
+   </td>
+   <td>99.07%
+   </td>
+  </tr>
+  <tr>
+   <td>TruthfulQA (0-shot)
+   </td>
+   <td>68.14
+   </td>
+   <td>66.32
+   </td>
+   <td>97.32%
+   </td>
+  </tr>
+  <tr>
+   <td><strong>Average</strong>
+   </td>
+   <td><strong>79.15</strong>
+   </td>
+   <td><strong>78.47</strong>
+   </td>
+   <td><strong>99.14%</strong>
+   </td>
+  </tr>
+</table>