<h3 id="transfer-from-huggingface">Transfer from huggingface</h3><a target="_blank" rel="ugc" href="https://huggingface.co/alibaba-pai/Wan2.2-Fun-Reward-LoRAs/tree/main">https://huggingface.co/alibaba-pai/Wan2.2-Fun-Reward-LoRAs/tree/main</a>sample of workflow: <a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/1966777795378655234?inviteCode=rh-v1213">https://www.runninghub.ai/post/1966777795378655234</a><h2 id="introduction">Introduction</h2>We explore the Reward Backpropagation technique <a target="_blank" rel="ugc" href="https://huggingface.co/alibaba-pai/Wan2.2-Fun-Reward-LoRAs#ref1">1</a> <a target="_blank" rel="ugc" href="https://huggingface.co/alibaba-pai/Wan2.2-Fun-Reward-LoRAs#ref2">2</a> to optimized the generated videos by <a target="_blank" rel="ugc" href="https://github.com/aigc-apps/VideoX-Fun">Wan2.2-Fun</a> for better alignment with human preferences. We provide the following pre-trained models (i.e. LoRAs) along with <a target="_blank" rel="ugc" href="https://github.com/aigc-apps/VideoX-Fun/blob/main/scripts/wan2.2_fun/train_reward_lora.py">the training script</a>. You can use these LoRAs to enhance the corresponding base model as a plug-in or train your own reward LoRA.For more details, please refer to our <a target="_blank" rel="ugc" href="https://github.com/aigc-apps/VideoX-Fun">GitHub repo</a>.<span style="background-color:rgb(11, 15, 25);color:rgb(179, 188, 201);font-family:&quot;Source Sans Pro&quot;, ui-sans-serif, system-ui, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;, &quot;Segoe UI Symbol&quot;, &quot;Noto Color Emoji&quot;;font-size:14.7px">Official MPS reward LoRA (<code>rank=128</code><span style="background-color:rgb(11, 15, 25);color:rgb(179, 188, 201);font-family:&quot;Source Sans Pro&quot;, ui-sans-serif, system-ui, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;, &quot;Segoe UI Symbol&quot;, &quot;Noto Color Emoji&quot;;font-size:14.7px"> and <code>network_alpha=64</code><span style="background-color:rgb(11, 15, 25);color:rgb(179, 188, 201);font-family:&quot;Source Sans Pro&quot;, ui-sans-serif, system-ui, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;, &quot;Segoe UI Symbol&quot;, &quot;Noto Color Emoji&quot;;font-size:14.7px">) for Wan2.2-Fun-A14B-InP (low noise). It is trained with a batch size of 8 for 4,500 steps.

Files with SHA256: 0e6ac56c89068386b002b18ab4542207e2c0c431f3c4ac367ae64fece85dd680