CUDA error when sampling

[vienduong88]

Hi, I have this error when running the t2v workflow. Thanks.

0 models unloaded.
loaded partially; 0.00 MB usable, 0.00 MB loaded, 1350.32 MB offloaded, 5219.06 MB buffer reserved, lowvram patches: 0
Found quantization metadata version 1
Detected mixed precision quantization
Using mixed precision operations
model weight dtype torch.bfloat16, manual cast: torch.bfloat16
model_type FLUX
unet unexpected: ['audio_embeddings_connector.learnable_registers', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.k_norm.weight', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.q_norm.weight', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.to_k.bias', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.to_k.weight', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.to_out.0.bias', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.to_out.0.weight', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.to_q.bias', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.to_q.weight', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.to_v.bias', 'audio_embeddings_connector.transformer_1d_blocks.0.attn1.to_v.weight', 'audio_embeddings_connector.transformer_1d_blocks.0.ff.net.0.proj.bias', 'audio_embeddings_connector.transformer_1d_blocks.0.ff.net.0.proj.weight', 'audio_embeddings_connector.transformer_1d_blocks.0.ff.net.2.bias', 'audio_embeddings_connector.transformer_1d_blocks.0.ff.net.2.weight', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.k_norm.weight', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.q_norm.weight', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.to_k.bias', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.to_k.weight', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.to_out.0.bias', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.to_out.0.weight', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.to_q.bias', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.to_q.weight', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.to_v.bias', 'audio_embeddings_connector.transformer_1d_blocks.1.attn1.to_v.weight', 'audio_embeddings_connector.transformer_1d_blocks.1.ff.net.0.proj.bias', 'audio_embeddings_connector.transformer_1d_blocks.1.ff.net.0.proj.weight', 'audio_embeddings_connector.transformer_1d_blocks.1.ff.net.2.bias', 'audio_embeddings_connector.transformer_1d_blocks.1.ff.net.2.weight', 'video_embeddings_connector.learnable_registers', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.k_norm.weight', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.q_norm.weight', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.to_k.bias', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.to_k.weight', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.to_out.0.bias', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.to_out.0.weight', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.to_q.bias', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.to_q.weight', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.to_v.bias', 'video_embeddings_connector.transformer_1d_blocks.0.attn1.to_v.weight', 'video_embeddings_connector.transformer_1d_blocks.0.ff.net.0.proj.bias', 'video_embeddings_connector.transformer_1d_blocks.0.ff.net.0.proj.weight', 'video_embeddings_connector.transformer_1d_blocks.0.ff.net.2.bias', 'video_embeddings_connector.transformer_1d_blocks.0.ff.net.2.weight', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.k_norm.weight', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.q_norm.weight', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.to_k.bias', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.to_k.weight', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.to_out.0.bias', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.to_out.0.weight', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.to_q.bias', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.to_q.weight', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.to_v.bias', 'video_embeddings_connector.transformer_1d_blocks.1.attn1.to_v.weight', 'video_embeddings_connector.transformer_1d_blocks.1.ff.net.0.proj.bias', 'video_embeddings_connector.transformer_1d_blocks.1.ff.net.0.proj.weight', 'video_embeddings_connector.transformer_1d_blocks.1.ff.net.2.bias', 'video_embeddings_connector.transformer_1d_blocks.1.ff.net.2.weight']
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
no CLIP/text encoder weights in checkpoint, the text encoder model will not be loaded.
Requested to load LTXAV
!!! Exception during processing !!! CUDA error: invalid argument
Search for `cudaErrorInvalidValue' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information.
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

Traceback (most recent call last):
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\execution.py", line 518, in execute
    output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
                                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\execution.py", line 329, in get_output_data
    return_values = await _async_map_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\execution.py", line 303, in _async_map_node_over_list
    await process_inputs(input_dict, i)
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\execution.py", line 291, in process_inputs
    result = f(**inputs)
             ^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy_api\internal\__init__.py", line 149, in wrapped_func
    return method(locked_class, **inputs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy_api\latest\_io.py", line 1570, in EXECUTE_NORMALIZED
    to_return = cls.execute(*args, **kwargs)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy_extras\nodes_custom_sampler.py", line 950, in execute
    samples = guider.sample(noise.generate_noise(latent), latent_image, sampler, sigmas, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=noise.seed)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\samplers.py", line 1050, in sample
    output = executor.execute(noise, latent_image, sampler, sigmas, denoise_mask, callback, disable_pbar, seed, latent_shapes=latent_shapes)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\patcher_extension.py", line 112, in execute
    return self.original(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\samplers.py", line 984, in outer_sample
    self.inner_model, self.conds, self.loaded_models = comfy.sampler_helpers.prepare_sampling(self.model_patcher, noise.shape, self.conds, self.model_options)
                                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\sampler_helpers.py", line 130, in prepare_sampling
    return executor.execute(model, noise_shape, conds, model_options=model_options, force_full_load=force_full_load)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\patcher_extension.py", line 112, in execute
    return self.original(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\sampler_helpers.py", line 138, in _prepare_sampling
    comfy.model_management.load_models_gpu([model] + models, memory_required=memory_required + inference_memory, minimum_memory_required=minimum_memory_required + inference_memory, force_full_load=force_full_load)
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\model_management.py", line 674, in load_models_gpu
    free_memory(total_memory_required[device] * 1.1 + extra_mem, device)
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\model_management.py", line 606, in free_memory
    if current_loaded_models[i].model_unload(memory_to_free):
       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\model_management.py", line 532, in model_unload
    self.model.detach(unpatch_weights)
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\model_patcher.py", line 989, in detach
    self.unpatch_model(self.offload_device, unpatch_weights=unpatch_all)
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\ComfyUI\comfy\model_patcher.py", line 857, in unpatch_model
    self.model.to(device_to)
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1371, in to
    return self._apply(convert)
           ^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 930, in _apply
    module._apply(fn)
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 930, in _apply
    module._apply(fn)
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 930, in _apply
    module._apply(fn)
  [Previous line repeated 4 more times]
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 957, in _apply
    param_applied = fn(param)
                    ^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1357, in convert
    return t.to(
           ^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\bitsandbytes\nn\modules.py", line 352, in to
    super().to(device=device, dtype=dtype, non_blocking=non_blocking),
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\DuongPC\AI\temp\ComfyUI_Easy\ComfyUI-Easy-Install\python_embeded\Lib\site-packages\bitsandbytes\nn\modules.py", line 402, in __torch_function__
    return super().__torch_function__(func, types, args, kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
torch.AcceleratorError: CUDA error: invalid argument
Search for `cudaErrorInvalidValue' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information.
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

[andrewladd]

I kept getting the same error. Adding these two arguments to my start script got it "running": --lowvram --disable-smart-memory, but... it's not "fast". I'm staring at 0/20 sample steps. But it's "running". Most of the other provided things didn't work, such as setting the VRAM number. --novram might have worked, but it took even longer to encode the prompts. I'm at a loss how people with even less compute than me are getting it to work much faster, but what can we do, I guess.

Edit: Okay, it seems like it's definitely the LTX-provided CLIP loader. Using the comfy-core-provided Clip loader ("LTXV Audio Text Encoder Loader"), and the Comfy Group-provided gemma, actually got it to both run and sample (albeit incredibly slow step 3/20 @ [03:06<17:31, 61.84s/it])

[dpmm99]

Same error, specifically from the LTXVImgToVideoInplace node and on a RTX 4060 Ti 16GB. andrewladd's steps didn't help. Was on CUDA 12.6, so I updated to 13.1 and reinstalled torch (2.9.1+cu130) and several other packages, but that didn't help. Windows 10, Python 3.11, all custom nodes and ComfyUI up-to-date, using --reserve-vram 0.5, resolution set to 640x352, bnb-4bit Gemma quant, Live Preview set to "none", with this change to comfy/ldm/lightricks/embeddings_connector.py:

hidden_states = torch.cat((hidden_states, learnable_registers[hidden_states.shape[1]:].unsqueeze(0).repeat(hidden_states.shape[0], 1, 1)), dim=1)

->

hidden_states = torch.cat((hidden_states, learnable_registers[hidden_states.shape[1]:].unsqueeze(0).repeat(hidden_states.shape[0], 1, 1).to(hidden_states.device)), dim=1)

Edit: I found a fix that works for me, just a command line parameter, --disable-pinned-memory: https://www.reddit.com/r/comfyui/comments/1pqh9qg/comment/nuu7vs0/ However, I did have Claude mess with some of the code a bit along the way, so I'll have to come back tomorrow and check if that alone fixes it.