LightX2V
LightX2V copied to clipboard
ModelTC
Light Video Generation Inference Framework
Great work! I'm wondering if you have any plan to support the recently released LongCat-Video-Avatar model from Meituan? Thanks.
### Description ring parallel can't work in wan2.2, RingAttnWeight.apply parameter img_qkv_len should be changed to slice_qkv_len. ### Steps to Reproduce ``` pipe = LightX2VPipeline( model_path="/root/wan22", model_cls="wan2.2_moe_distill", task="i2v", low_noise_original_ckpt="/root/wan22/low_noise_model/wan2.2_i2v_A14b_low_noise_scaled_fp8_e4m3_lightx2v_4step.safetensors", high_noise_original_ckpt="/root/wan22/high_noise_model/wan2.2_i2v_A14b_high_noise_scaled_fp8_e4m3_lightx2v_4step.safetensors", #...
如题，我想用自己的数据集训练一个加速lora，但是没有找到很明确的方法和代码。请问我要怎么做，有可用的训练框架或者代码吗，感谢
https://github.com/user-attachments/assets/72576806-3e60-463b-b65e-6f8815a518b8 The result looks very poor. Here is my code. Does anyone else have this problem? ` """ Wan2.2 distilled model image-to-video generation example. This example demonstrates how to use...
**wan_moe_i2v_distill.json：** { "infer_steps": 4, "target_video_length": 81, "text_len": 512, "target_height": 720, "target_width": 1280, "self_attn_1_type": "sage_attn2", "cross_attn_1_type": "sage_attn2", "cross_attn_2_type": "sage_attn2", "dit_quantized": true, "dit_quant_scheme": "fp8-vllm" "high_noise_original_ckpt": "/mnt/cache1/LightX2V/Wan2.2-Distill-Models/high_noise_model/wan2.2_i2v_A14b_high_noise_scaled_fp8_e4m3_lightx2v_4step.safetensors", "low_noise_original_ckpt": "/mnt/cache1/LightX2V/Wan2.2-Distill-Models/low_noise_model/wan2.2_i2v_A14b_low_noise_scaled_fp8_e4m3_lightx2v_4step.safetensors", "sample_guide_scale": [ 7.0, 11.0...
t2v only? released 4-step distilled models for HunyuanVideo-1.5 is only t2v ?
### Description Briefly describe the bug you encountered. I'm comparing the Diffusers implementation by the Authors to your Implementation for WAN2.2. So the Non distilled version compared to the original...
Hi, in the comfyui workflows lot's of people use a KSampler to automatically determine the split between high and low noise models. Is there something similar in Lightx2v? How can...