cyan_s comments

Results 9 comments of


                                            cyan_s

Finetune img2video based on T2V model

这里的噪声强度在什么范围会比较合适呢？我用的0.02前半部分会有些动态，后面的偏向于静态

2B的模型可以支持微调高分辨吗，报错RuntimeError: integer out of range

谢谢

2B的模型可以支持微调高分辨吗，报错RuntimeError: integer out of range

在微调720,480分辨率模型时候，出现了下面问题，请问这是什么情况呢？ [2024-09-13 12:06:43,971] [INFO] [RANK 0] ---------------------------------------------------------------------------------------------------- [2024-09-13 12:06:43,972] [INFO] [RANK 0] ----------------------------------------------------------------------------------------------- [2024-09-13 12:06:43,972] [INFO] [RANK 0] validation loss at iteration 900 | loss: 9.150388E-01 | PPL: 2.496872E+00...

两张参考图的情况下，这里是不是有问题？应该有下标索引image_vae[0]么？

这块没有理解，这是在做什么当参考图数量为2时， video_condition = 图一+图二* 4+ 零*（num_frames-5）当参考图数量为3时，vae_repeat的情况下，video_condition = 图一+图二* 4+图三* 4+零*（num_frames-9）图一不用repeat么？

OSError: Error no file named model_index.json found in directory

这个json似乎不对

OSError: Error no file named model_index.json found in directory

{ "_class_name": "A2Pipeline", "_diffusers_version": "0.33.0.dev0", "transformer": ["models.transformer_a2", "A2Model"], "vae": ["diffusers", "AutoencoderKLWan"], "image_encoder": ["transformers", "CLIPVisionModelWithProjection"], "text_encoder": ["transformers", "UMT5EncoderModel"], "scheduler": ["diffusers", "UniPCMultistepScheduler"], "tokenizer": ["transformers", "T5TokenizerFast"], "image_processor": ["transformers", "CLIPImageProcessor"] }或许应该是这样

Batch Mode + Maintain Aspect Ratio + Multi-GPU Random Seed + Fixed Multi-GPU CuSolver Error + Fixed 20-Min Load Time + Video Input & Multiple Prompts

感谢分享，官方df的不支持竖屏，多卡跑也加载不动，单卡跑的df 宽屏跑了四个小时257帧

Batch Mode + Maintain Aspect Ratio + Multi-GPU Random Seed + Fixed Multi-GPU CuSolver Error + Fixed 20-Min Load Time + Video Input & Multiple Prompts

尝试跑了下，会报错：