Update README.md

README.md

@@ -32,7 +32,7 @@ base_model: meta-llama/Meta-Llama-3.1-405B-Instruct
 - **Model Developers:** Neural Magic
 
 Quantized version of [Meta-Llama-3.1-405B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct).
-It achieves an average score of 86.47 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 86.63.
+It achieves scores within 1% of the scores of the unquantized model for MMLU, ARC-Challenge, GSM-8k, Hellaswag, Winogrande, and TruthfulQA.
 
 ### Model Optimizations
 
@@ -130,8 +130,9 @@ model.save_pretrained("Meta-Llama-3.1-405B-Instruct-quantized.w4a16")
 
 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 ARC-Challenge, GSM-8K, and MMLU 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).
+This version of the lm-evaluation-harness includes versions of ARC-Challenge, GSM-8K, and MMLU that match the prompting style of [Meta-Llama-3.1-Instruct-evals](https://huggingface.co/datasets/meta-llama/Meta-Llama-3.1-405B-Instruct-evals).
 
+**Note:** Results have been updated after Meta modified the chat template.
 
 ### Accuracy
 
@@ -144,27 +145,27 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td><strong>Meta-Llama-3.1-405B-Instruct-quantized.w4a16 (this model)</strong>
    </td>
-   <td><strong>Recovery (this model) </strong>
+   <td><strong>Recovery</strong>
    </td>
   </tr>
   <tr>
    <td>MMLU (5-shot)
    </td>
-   <td>86.25
+   <td>87.38
    </td>
-   <td>86.17
+   <td>87.22
    </td>
-   <td>99.90%
+   <td>99.8%
    </td>
   </tr>
   <tr>
    <td>ARC Challenge (0-shot)
    </td>
-   <td>96.93
+   <td>94.97
    </td>
-   <td>95.3
+   <td>95.31
    </td>
-   <td>98.31%
+   <td>100.4%
    </td>
   </tr>
   <tr>
@@ -172,9 +173,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>96.44
    </td>
-   <td>96.05
+   <td>96.29
    </td>
-   <td>99.59%
+   <td>99.8%
    </td>
   </tr>
   <tr>
@@ -184,7 +185,7 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>88.27
    </td>
-   <td>99.93%
+   <td>99.9%
    </td>
   </tr>
   <tr>
@@ -192,9 +193,9 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>87.21
    </td>
-  <td>87.76
+   <td>87.37
    </td>
-   <td>100.63%
+   <td>100.2%
    </td>
   </tr>
   <tr>
@@ -202,19 +203,19 @@ This version of the lm-evaluation-harness includes versions of ARC-Challenge, GS
    </td>
    <td>64.64
    </td>
-   <td>65.27
+   <td>65.26
    </td>
-   <td>100.97%
+   <td>101.0%
    </td>
   </tr>
   <tr>
    <td><strong>Average</strong>
    </td>
-   <td><strong>86.63</strong>
+   <td><strong>86.75</strong>
    </td>
-   <td><strong>86.47</strong>
+   <td><strong>86.76</strong>
    </td>
-   <td><strong>99.81%</strong>
+   <td><strong>100.0%</strong>
    </td>
   </tr>
 </table>
@@ -227,7 +228,7 @@ The results were obtained using the following commands:
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w4a16",dtype=auto,add_bos_token=True,max_model_len=4096,max_gen_toks=10,tensor_parallel_size=8 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w4a16",dtype=auto,max_model_len=4096,max_gen_toks=10,tensor_parallel_size=8 \
   --tasks mmlu_llama_3.1_instruct \
   --apply_chat_template \
   --fewshot_as_multiturn \
@@ -239,7 +240,7 @@ lm_eval \
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w4a16",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=8 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w4a16",dtype=auto,max_model_len=4096,tensor_parallel_size=8 \
   --tasks arc_challenge_llama_3.1_instruct \
   --apply_chat_template \
   --num_fewshot 0 \
@@ -250,7 +251,7 @@ lm_eval \
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w4a16",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=8 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-405B-Instruct-quantized.w4a16",dtype=auto,max_model_len=4096,tensor_parallel_size=8 \
   --tasks gsm8k_cot_llama_3.1_instruct \
   --apply_chat_template \
   --fewshot_as_multiturn \