Update README.md

README.md

@@ -33,7 +33,7 @@ base_model: meta-llama/Meta-Llama-3.1-70B-Instruct
 - **Model Developers:** Neural Magic
 
 Quantized version of [Meta-Llama-3.1-70B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-70B-Instruct).
-It achieves scores within 1% of the scores of the unquantized model for MMLU, ARC-Challenge, GSM-8k, Hellaswag, Winogrande and TruthfulQA.
+It achieves scores within 1.2% of the scores of the unquantized model for MMLU, ARC-Challenge, GSM-8k, Hellaswag, Winogrande and TruthfulQA.
 
 ### Model Optimizations
 
@@ -136,6 +136,8 @@ The model was evaluated on MMLU, ARC-Challenge, GSM-8K, Hellaswag, Winogrande an
 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-3.1-70B-Instruct-evals).
 
+**Note:** Results have been updated after Meta modified the chat template.
+
 ### Accuracy
 
 #### Open LLM Leaderboard evaluation scores
@@ -153,9 +155,9 @@ This version of the lm-evaluation-harness includes versions of MMLU, ARC-Challen
   <tr>
    <td>MMLU (5-shot)
    </td>
-   <td>83.88
+   <td>83.94
    </td>
-   <td>83.65
+   <td>83.71
    </td>
    <td>99.7%
    </td>
@@ -163,71 +165,71 @@ This version of the lm-evaluation-harness includes versions of MMLU, ARC-Challen
   <tr>
    <td>MMLU (CoT, 0-shot)
    </td>
-   <td>85.74
+   <td>86.23
    </td>
-   <td>85.41
+   <td>85.81
    </td>
-   <td>99.6%
+   <td>99.5%
    </td>
   </tr>
   <tr>
    <td>ARC Challenge (0-shot)
    </td>
-   <td>93.26
+   <td>93.34
    </td>
-   <td>93.26
+   <td>93.09
    </td>
-   <td>100.0%
+   <td>99.7%
    </td>
   </tr>
   <tr>
    <td>GSM-8K (CoT, 8-shot, strict-match)
    </td>
-   <td>93.10
+   <td>95.38
    </td>
-   <td>93.25
+   <td>94.24
    </td>
-   <td>100.2%
+   <td>98.8%
    </td>
   </tr>
   <tr>
    <td>Hellaswag (10-shot)
    </td>
-   <td>86.40
+   <td>86.66
    </td>
-   <td>86.28
+   <td>86.19
    </td>
-   <td>99.9%
+   <td>99.5%
    </td>
   </tr>
   <tr>
    <td>Winogrande (5-shot)
    </td>
-   <td>85.00
+   <td>85.32
    </td>
    <td>85.00
    </td>
-   <td>100.0%
+   <td>99.6%
    </td>
   </tr>
   <tr>
    <td>TruthfulQA (0-shot, mc2)
    </td>
-   <td>59.83
+   <td>60.65
    </td>
-   <td>60.88
+   <td>60.69
    </td>
-   <td>101.8%
+   <td>100.1%
    </td>
   </tr>
   <tr>
    <td><strong>Average</strong>
    </td>
-   <td><strong>83.89</strong>
+   <td><strong>84.50</strong>
    </td>
-   <td><strong>83.96</strong>
+   <td><strong>84.10</strong>
    </td>
-   <td><strong>100.2%</strong>
+   <td><strong>99.6%</strong>
    </td>
   </tr>
 </table>
@@ -240,7 +242,7 @@ The results were obtained using the following commands:
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w8a8",dtype=auto,add_bos_token=True,max_model_len=3850,max_gen_toks=10,tensor_parallel_size=1 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w8a8",dtype=auto,max_model_len=3850,max_gen_toks=10,tensor_parallel_size=1 \
   --tasks mmlu_llama_3.1_instruct \
   --fewshot_as_multiturn \
   --apply_chat_template \
@@ -252,7 +254,7 @@ lm_eval \
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w8a8",dtype=auto,add_bos_token=True,max_model_len=4064,max_gen_toks=1024,tensor_parallel_size=1 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w8a8",dtype=auto,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 \
@@ -263,7 +265,7 @@ lm_eval \
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w8a8",dtype=auto,add_bos_token=True,max_model_len=3940,max_gen_toks=100,tensor_parallel_size=1 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w8a8",dtype=auto,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 \
@@ -274,7 +276,7 @@ lm_eval \
 ```
 lm_eval \
   --model vllm \
-  --model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w8a8",dtype=auto,add_bos_token=True,max_model_len=4096,max_gen_toks=1024,tensor_parallel_size=1 \
+  --model_args pretrained="neuralmagic/Meta-Llama-3.1-70B-Instruct-quantized.w8a8",dtype=auto,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 \