a bytes-like object is required, not 'NoneType'

[JakeCirino]

ComfyUI Error Report

Error Details

• Node ID: 2
• Node Type: STATIC_TRT_MODEL_CONVERSION
• Exception Type: TypeError
• Exception Message: a bytes-like object is required, not 'NoneType'

Stack Trace

  File "/app/execution.py", line 327, in execute
    output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "/app/execution.py", line 202, in get_output_data
    return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "/app/execution.py", line 174, in _map_node_over_list
    process_inputs(input_dict, i)

  File "/app/execution.py", line 163, in process_inputs
    results.append(getattr(obj, func)(**inputs))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "/app/custom_nodes/ComfyUI_TensorRT/tensorrt_convert.py", line 627, in convert
    return super()._convert(
           ^^^^^^^^^^^^^^^^^

  File "/app/custom_nodes/ComfyUI_TensorRT/tensorrt_convert.py", line 384, in _convert
    f.write(serialized_engine)

System Information

• ComfyUI Version: 0.3.27
• Arguments: main.py --listen 0.0.0.0
• OS: posix
• Python Version: 3.12.3 (main, Feb 4 2025, 14:48:35) [GCC 13.3.0]
• Embedded Python: false
• PyTorch Version: 2.6.0+cu124

Devices

• Name: cuda:0 Tesla V100-PCIE-16GB : cudaMallocAsync
  • Type: cuda
  • VRAM Total: 16935682048
  • VRAM Free: 16350838784
  • Torch VRAM Total: 33554432
  • Torch VRAM Free: 25034752

Logs

2025-03-29T22:08:09.119165 - Checkpoint files will always be loaded safely.
2025-03-29T22:08:10.825141 - Total VRAM 16151 MB, total RAM 112686 MB
2025-03-29T22:08:10.825226 - pytorch version: 2.6.0+cu124
2025-03-29T22:08:10.825612 - Set vram state to: NORMAL_VRAM
2025-03-29T22:08:10.825901 - Device: cuda:0 Tesla V100-PCIE-16GB : cudaMallocAsync
2025-03-29T22:08:12.208600 - Using pytorch attention
2025-03-29T22:08:14.280069 - ComfyUI version: 0.3.27
2025-03-29T22:08:14.283359 - ****** User settings have been changed to be stored on the server instead of browser storage. ******
2025-03-29T22:08:14.283417 - ****** For multi-user setups add the --multi-user CLI argument to enable multiple user profiles. ******
2025-03-29T22:08:14.284077 - ComfyUI frontend version: 1.14.6
2025-03-29T22:08:14.284937 - [Prompt Server] web root: /usr/local/lib/python3.12/dist-packages/comfyui_frontend_package/static
2025-03-29T22:08:14.891447 - 
Import times for custom nodes:
2025-03-29T22:08:14.891542 -    0.0 seconds: /app/custom_nodes/websocket_image_save.py
2025-03-29T22:08:14.891595 -    0.1 seconds: /app/custom_nodes/ComfyUI_TensorRT
2025-03-29T22:08:14.891640 - 
2025-03-29T22:08:14.898721 - Starting server

2025-03-29T22:08:14.899020 - To see the GUI go to: http://0.0.0.0:8188
2025-03-29T22:08:14.899083 - Instrumentation key provided, emitting queue depth metric to azure
2025-03-29T22:08:15.214280 - Queue monitoring started
2025-03-29T22:09:13.334229 - got prompt
2025-03-29T22:09:13.446817 - model weight dtype torch.float16, manual cast: None
2025-03-29T22:09:13.447635 - model_type EPS
2025-03-29T22:09:13.754403 - Using pytorch attention in VAE
2025-03-29T22:09:13.755997 - Using pytorch attention in VAE
2025-03-29T22:09:13.943397 - VAE load device: cuda:0, offload device: cpu, dtype: torch.float32
2025-03-29T22:09:14.021409 - CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
2025-03-29T22:09:14.157284 - Requested to load BaseModel
2025-03-29T22:09:14.425388 - loaded completely 9.5367431640625e+25 1639.406135559082 True
2025-03-29T22:09:14.722331 - /app/comfy/ldm/modules/attention.py:447: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if SDP_BATCH_LIMIT >= b:
2025-03-29T22:09:14.780642 - /app/comfy/ldm/modules/diffusionmodules/openaimodel.py:132: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  assert x.shape[1] == self.channels
2025-03-29T22:09:15.107188 - /app/comfy/ldm/modules/diffusionmodules/openaimodel.py:90: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  assert x.shape[1] == self.channels
2025-03-29T22:09:29.260691 - No timing cache found; Initializing a new one.2025-03-29T22:09:29.260727 - 
2025-03-29T22:09:29.262557 - !!! Exception during processing !!! a bytes-like object is required, not 'NoneType'
2025-03-29T22:09:29.264426 - Traceback (most recent call last):
  File "/app/execution.py", line 327, in execute
    output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/execution.py", line 202, in get_output_data
    return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/execution.py", line 174, in _map_node_over_list
    process_inputs(input_dict, i)
  File "/app/execution.py", line 163, in process_inputs
    results.append(getattr(obj, func)(**inputs))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/custom_nodes/ComfyUI_TensorRT/tensorrt_convert.py", line 627, in convert
    return super()._convert(
           ^^^^^^^^^^^^^^^^^
  File "/app/custom_nodes/ComfyUI_TensorRT/tensorrt_convert.py", line 384, in _convert
    f.write(serialized_engine)
TypeError: a bytes-like object is required, not 'NoneType'

2025-03-29T22:09:29.265531 - Prompt executed in 15.93 seconds
2025-03-29T22:09:38.229828 - got prompt
2025-03-29T22:09:38.232402 - Requested to load BaseModel
2025-03-29T22:09:38.511050 - loaded completely 9.5367431640625e+25 1639.406135559082 True
2025-03-29T22:09:49.916097 - No timing cache found; Initializing a new one.2025-03-29T22:09:49.916131 - 
2025-03-29T22:09:49.917462 - !!! Exception during processing !!! a bytes-like object is required, not 'NoneType'
2025-03-29T22:09:49.918632 - Traceback (most recent call last):
  File "/app/execution.py", line 327, in execute
    output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/execution.py", line 202, in get_output_data
    return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/execution.py", line 174, in _map_node_over_list
    process_inputs(input_dict, i)
  File "/app/execution.py", line 163, in process_inputs
    results.append(getattr(obj, func)(**inputs))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/custom_nodes/ComfyUI_TensorRT/tensorrt_convert.py", line 627, in convert
    return super()._convert(
           ^^^^^^^^^^^^^^^^^
  File "/app/custom_nodes/ComfyUI_TensorRT/tensorrt_convert.py", line 384, in _convert
    f.write(serialized_engine)
TypeError: a bytes-like object is required, not 'NoneType'

2025-03-29T22:09:49.919951 - Prompt executed in 11.69 seconds

Attached Workflow

Please make sure that workflow does not contain any sensitive information such as API keys or passwords.

{"id":"f61e5314-16b8-419f-b32a-4d4b18ba0b52","revision":0,"last_node_id":2,"last_link_id":1,"nodes":[{"id":1,"type":"CheckpointLoaderSimple","pos":[681.5532836914062,263.5812683105469],"size":[315,98],"flags":{},"order":0,"mode":0,"inputs":[],"outputs":[{"localized_name":"MODEL","name":"MODEL","type":"MODEL","links":[1]},{"localized_name":"CLIP","name":"CLIP","type":"CLIP","links":null},{"localized_name":"VAE","name":"VAE","type":"VAE","links":null}],"properties":{"Node name for S&R":"CheckpointLoaderSimple"},"widgets_values":["realisticVisionV51_v51VAE.safetensors"]},{"id":2,"type":"STATIC_TRT_MODEL_CONVERSION","pos":[1108.68310546875,270.84124755859375],"size":[340.20001220703125,178],"flags":{},"order":1,"mode":0,"inputs":[{"localized_name":"model","name":"model","type":"MODEL","link":1}],"outputs":[],"properties":{"Node name for S&R":"STATIC_TRT_MODEL_CONVERSION"},"widgets_values":["tensorrt/realism",1,768,512,1,14]}],"links":[[1,1,0,2,0,"MODEL"]],"groups":[],"config":{},"extra":{"ds":{"scale":0.683013455365071,"offset":[-6.929903795422533,22.632130376695077]}},"version":0.4}

Additional Context

I am getting this error when trying to generate a static TensorRT model on an Nvidia V100. I am able to run this workflow on an identical docker container running on an Nvidia T4.

I am using nvcr.io/nvidia/cuda:12.8.1-cudnn-runtime-ubuntu24.04 as my base image

[JakeCirino]

I was able to get more verbose logging from tensorrt by modifying line 24 of tensorrt_loader.py:

logger = trt.Logger(trt.Logger.VERBOSE) logger.min_severity = trt.Logger.VERBOSE

After which, I was able to locate the following error:

IBuilder::buildSerializedNetwork: Error Code 9: API Usage Error (Target GPU SM 70 is not supported by this TensorRT release.)

So after an immense amount of googling to find any relevant info, I discovered that TensorRT > 10.5 no longer supports the Volta architecture.

I was able to downgrade the tensorrt 10.4 and the model generation worked!

However, for some reason running an equivalent workflow with a normal model/tensorrt model, the tensorrt model was about 0.7s slower than the regular model.

I am assuming trying to get use of tensor cores on anything with older architecture is a lost cause on this point?