[doc] refine performance image size (#39)

README.md

@@ -49,11 +49,7 @@ Please refer to the [documentation](https://chatlearn.readthedocs.io/en/latest/)
 
 We compared the RLHF training throughput of models with different parameter scales, adopting an N+N model configuration where both the Policy model and the Reward model have the same number of parameters. We benchmarked against DeepSpeed-Chat and OpenRLHF with 7B and 70B model configurations. For the 8 GPU setup with a 7B+7B scale, we achieved a 115% speedup; for the 32 GPU setup with a 70B+70B scale, the speedup was 208%. The larger the scale, the more pronounced the acceleration effect becomes. Additionally, ChatLearn can support even larger-scale alignment training, such as at a 300B+300B scale.
 
-<p align="center">
-  <picture>
-    <img alt="compare perf" src="docs/images/perf.png" width=50%>
-  </picture>
-</p>
+![Compare Performance](docs/images/perf.png)
 
 Note: The performance of DeepSpeed-Chat and OpenRLHF has already been optimized.
 

README_CN.md

@@ -45,11 +45,8 @@ ChatLearn的特点如下:
 
 我们比较了不同参数量规模模型的 RLHF 训练吞吐量，我们采取 N+N 的模型配置，即 Policy 模型和 Reward 模型采用相同大小的参数量。我们和 DeepSpeed-Chat、OpenRLHF 对比了 7B 和 70B 的模型配置，在 8 GPUs 7B+7B 规模，有 115% 的加速，在 32 GPUs 70B+70B 规模，有 208% 的加速。规模越大，加速效果越明显。同时ChatLearn还能支持更大规模的 Alignment 训练，例如：300B+300B 规模。
 
-<p align="center">
-  <picture>
-    <img alt="compare perf" src="docs/images/perf.png" width=50%>
-  </picture>
-</p>
+
+![Compare Performance](docs/images/perf.png)
 
 注：DeepSpeed-Chat和OpenRLHF性能已经优化过。