module 'triton' has no attribute 'language'

[NguyenThanhHa288]

I'm getting the error message "AttributeError: module 'triton' has no attribute 'language'" when I update the triton library from version 2.47.0 to 2.48.0.

[rmccorm4]

Hi @NguyenThanhHa288, can you share more details on how you're installing and what code you're writing that is generating this error?

Please provide the minimal steps to reproduce this error.

[msoftware]

Same issue here with torch-2.4.0

File "/home/michael/.local/lib/python3.10/site-packages/torch/_dynamo/init.py", line 2, in from . import convert_frame, eval_frame, resume_execution File "/home/michael/.local/lib/python3.10/site-packages/torch/_dynamo/convert_frame.py", line 40, in from . import config, exc, trace_rules File "/home/michael/.local/lib/python3.10/site-packages/torch/_dynamo/exc.py", line 11, in from .utils import counters File "/home/michael/.local/lib/python3.10/site-packages/torch/_dynamo/utils.py", line 1008, in common_constant_types.add(triton.language.dtype) AttributeError: module 'triton' has no attribute 'language'

[wwdok]

I got the same issue with @msoftware, this error comes from common_constant_types.add(triton.language.dtype),to reproduce this error, just execute from torchvision import transforms， my torchvision verison is 0.19.1

[rmccorm4]

Hi, this looks like use of OpenAI Triton, and not NVIDIA Triton Inference Server. Please raise an issue here: https://github.com/triton-lang/triton

[jaffe-fly]

same error code is

triton_server_url = 'xxxxxxx:8003'
triton_client = InferenceServerClient(url=triton_server_url, verbose=False, ssl=False)


if triton_client.is_model_ready(model_name):

    # Load the Triton Server model
    model = YOLO(f"http://{triton_server_url}/yolov8x", task="detect")

    # Run inference on the server
    results = model.predict(r"E:\code\images\val\frame_027940.PNG")
    for r in results:
        print(f"Detected {len(r)} objects in image")

error is

Traceback (most recent call last):
  File "E:\code\cards\api\test_triton.py", line 15, in <module>
    results = model.predict(r"E:\code\cards\train\datasets\221319\images\val\frame_027940.PNG")
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "E:\miniconda\envs\card\Lib\site-packages\ultralytics\engine\model.py", line 554, in predict
    return self.predictor.predict_cli(source=source) if is_cli else self.predictor(source=source, stream=stream)
                                                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "E:\miniconda\envs\card\Lib\site-packages\ultralytics\engine\predictor.py", line 168, in __call__
    return list(self.stream_inference(source, model, *args, **kwargs))  # merge list of Result into one
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "E:\miniconda\envs\card\Lib\site-packages\torch\utils\_contextlib.py", line 36, in generator_context
    response = gen.send(None)
               ^^^^^^^^^^^^^^
  File "E:\miniconda\envs\card\Lib\site-packages\ultralytics\engine\predictor.py", line 234, in stream_inference
    self.model.warmup(imgsz=(1 if self.model.pt or self.model.triton else self.dataset.bs, 3, *self.imgsz))
  File "E:\miniconda\envs\card\Lib\site-packages\ultralytics\nn\autobackend.py", line 642, in warmup
    import torchvision  # noqa (import here so torchvision import time not recorded in postprocess time)
    ^^^^^^^^^^^^^^^^^^
  File "E:\miniconda\envs\card\Lib\site-packages\torchvision\__init__.py", line 10, in <module>
    from torchvision import _meta_registrations, datasets, io, models, ops, transforms, utils  # usort:skip
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "E:\miniconda\envs\card\Lib\site-packages\torchvision\models\__init__.py", line 2, in <module>
    from .convnext import *
  File "E:\miniconda\envs\card\Lib\site-packages\torchvision\models\convnext.py", line 8, in <module>
    from ..ops.misc import Conv2dNormActivation, Permute
  File "E:\miniconda\envs\card\Lib\site-packages\torchvision\ops\__init__.py", line 23, in <module>
    from .poolers import MultiScaleRoIAlign
  File "E:\miniconda\envs\card\Lib\site-packages\torchvision\ops\poolers.py", line 10, in <module>
    from .roi_align import roi_align
  File "E:\miniconda\envs\card\Lib\site-packages\torchvision\ops\roi_align.py", line 7, in <module>
    from torch._dynamo.utils import is_compile_supported
  File "E:\miniconda\envs\card\Lib\site-packages\torch\_dynamo\__init__.py", line 2, in <module>
    from . import convert_frame, eval_frame, resume_execution
  File "E:\miniconda\envs\card\Lib\site-packages\torch\_dynamo\convert_frame.py", line 48, in <module>
    from . import config, exc, trace_rules
  File "E:\miniconda\envs\card\Lib\site-packages\torch\_dynamo\exc.py", line 12, in <module>
    from .utils import counters
  File "E:\miniconda\envs\card\Lib\site-packages\torch\_dynamo\utils.py", line 1066, in <module>
    common_constant_types.add(triton.language.dtype)
                              ^^^^^^^^^^^^^^^
AttributeError: module 'triton' has no attribute 'language'

torch                     2.4.1
torchvision               0.19.1
tqdm                      4.66.5
tritonclient              2.51.0

server_version 2.50.0

[jaffe-fly]

so sorry ,in my project,there is a dir named triton, its a bad dirname,after refactor it ,is ok