Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

update readme (#1141)

Wenhua Chengcde5160a0221
diff --git a/README.md b/README.md
index c2cc6c3e3..f69ce037e 100644
--- a/README.md
+++ b/README.md
@@ -31,7 +31,8 @@ See our [paper](https://arxiv.org/pdf/2309.05516) for more details. For usage in
## 🆕 What's New
-* [2025/12] The [AutoRoundV2 paper](http://arxiv.org/abs/2512.04746) is now publicly available on arXiv. Turn on `enable_alg_ext` and use the **AutoScheme** API for mixed-precision quantization to reproduce the results.
+* [2025/12] The [AutoRoundV2 paper](http://arxiv.org/abs/2512.04746) is now publicly available on arXiv. Turn on `enable_alg_ext` and use the **AutoScheme** API for mixed-precision quantization to reproduce the results. See the evaluation [notes](./docs/alg_202508.md) for LLaMA models
+ for additional guidance.
* [2025/11] AutoRound has landed in **LLM-Compressor**: [*Usage*](https://github.com/vllm-project/llm-compressor/tree/main/examples/autoround/README.md).
diff --git a/docs/alg_202508.md b/docs/alg_202508.md
index 4c5a62520..14828e230 100644
--- a/docs/alg_202508.md
+++ b/docs/alg_202508.md
@@ -1,7 +1,7 @@
-We use **lm-eval** for evaluation. For LLaMA, we enabled `add_bos_token` and
-`removed @use_kernel_forward_from_hub("RMSNorm")`
-in [modeling_llama.py](https://github.com/huggingface/transformers/blob/main/src/transformers/models/llama/modeling_llama.py#L52C1-L52C40)
-to stabilize accuracy during evaluation. All other settings follow the default configurations of AutoRound and lm-eval.
+ If you are evaluating LLaMA models with recent versions of Transformers, please
+remove `@use_kernel_forward_from_hub("RMSNorm")` in [modeling_llama.py](https://github.com/huggingface/transformers/blob/main/src/transformers/models/llama/modeling_llama.py#L52C1-L52C40) and enable `add_bos_token`(this is set as default in AutoRound) in lm-eval to stabilize the accuracy. These adjustments affect the quantized model but not the BF16 model for the tasks evaluated in the AutoRoundv2 paper.
+
+All other settings follow the default configurations of AutoRound and lm-eval.
| Qwen3-8B W2G64 | Avg. | arc_challenge | hellaswag | gsm8k | lambada_openai | mmlu | mmlupro | truthfulqa_mc1 | winogrande |
|:------------------------------|:------:|:-------------:|:---------:|:------:|:--------------:|:------:|:-------:|:--------------:|:----------:|