AssertionError: context parallel size is not initialized #20
Description
I'm having problem:
[2025-03-13 08:34:22,169] [INFO] [RANK 0] merge layer 35 lora attention back to linear
INFO:sat:[RANK 0] merge layer 35 lora attention back to linear
[2025-03-13 08:34:22,170] [INFO] [RANK 0] merge layer 36 lora attention back to linear
INFO:sat:[RANK 0] merge layer 36 lora attention back to linear
[2025-03-13 08:34:22,170] [INFO] [RANK 0] merge layer 37 lora attention back to linear
INFO:sat:[RANK 0] merge layer 37 lora attention back to linear
[2025-03-13 08:34:22,171] [INFO] [RANK 0] merge layer 38 lora attention back to linear
INFO:sat:[RANK 0] merge layer 38 lora attention back to linear
[2025-03-13 08:34:22,172] [INFO] [RANK 0] merge layer 39 lora attention back to linear
INFO:sat:[RANK 0] merge layer 39 lora attention back to linear
[2025-03-13 08:34:22,173] [INFO] [RANK 0] merge layer 40 lora attention back to linear
INFO:sat:[RANK 0] merge layer 40 lora attention back to linear
[2025-03-13 08:34:22,174] [INFO] [RANK 0] merge layer 41 lora attention back to linear
INFO:sat:[RANK 0] merge layer 41 lora attention back to linear
merge lora of model.diffusion_model.mixins.lora
rank and world_size 0 1
load from /home/wrusr/genai_workspace/flash_env/workspace/FlashVideo/example.txt with 25
vis_270p_1080p_example
rank 0 processing 0
{'prompt': 'The video begins with a view of a cozy log cabin nestled among lush trees in a dense forest. A gentle plume of white smoke rises gracefully from the chimney, curling against the backdrop of towering pines. The windows emit a warm, soft glow, hinting at a comforting and inviting atmosphere within. The camera slowly pans from right to left, capturing the serene setting and emphasizing the harmony between the cabin and its natural surroundings.', 'negative_prompt': 'very low motion,, low motion,', 'num_frames': tensor([13])}
txt [445]
overwrite cfg scale in config of stage-1
############################## Sampling setting ##############################
Sampler: VPSDEDPMPP2MSampler
Discretization: ZeroSNRDDPMDiscretization
Guider: DynamicCFG
Sampling with VPSDEDPMPP2MSampler for 52 steps: 98%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▌ | 51/52 [01:33<00:01, 1.83s/it]
_CONTEXT_PARALLEL_SIZE: None
[rank0]: Traceback (most recent call last):
[rank0]: File "/home/wrusr/genai_workspace/flash_env/workspace/FlashVideo/flashvideo/dist_inf_text_file.py", line 390, in
[rank0]: sampling_main(args,
[rank0]: File "/home/wrusr/genai_workspace/flash_env/workspace/FlashVideo/flashvideo/dist_inf_text_file.py", line 308, in sampling_main
[rank0]: first_stage_samples = get_first_results(model, text, num_frames, H, W)
[rank0]: File "/home/wrusr/genai_workspace/flash_env/workspace/FlashVideo/flashvideo/dist_inf_text_file.py", line 124, in get_first_results
[rank0]: samples = decode(first_stage_model, latent)
[rank0]: File "/home/wrusr/genai_workspace/flash_env/workspace/FlashVideo/flashvideo/utils.py", line 81, in decode
[rank0]: recon = first_stage_model.decode(
[rank0]: File "/home/wrusr/genai_workspace/flash_env/workspace/FlashVideo/flashvideo/vae_modules/autoencoder.py", line 704, in decode
[rank0]: global_src_rank = get_context_parallel_group_rank() * self.cp_size
[rank0]: File "/home/wrusr/genai_workspace/flash_env/workspace/FlashVideo/flashvideo/sgm/util.py", line 71, in get_context_parallel_group_rank
[rank0]: assert _CONTEXT_PARALLEL_SIZE is not None, 'context parallel size is not initialized'
[rank0]: AssertionError: context parallel size is not initialized
How can I solve this problem?