Project test failing: transcribe test fails with core dump

[hmf]

Hello,

First and foremost, thanks for making this publicity available.

I am trying to experiment with the project using the git repo. My setup i as follows:

• Linux Ubuntu
• GPU: Quadro P1000 with 4GiB
• CUDA Version: 12.4
  • NVIDIA-SMI 550.120
  • Driver Version: 550.120
• Using a fork from this repository
• Using a branch from the original master
• Created and using a virtual environment
• Using Python 3.12.3
• Installed all libraries from all 3 requirements files
• Executed the test as indicated in the CONTRIBUTING.md file

The following tests are successful:

1. pytest -v tests/test_utils.py
2. pytest -v tests/test_tokenizer.py

The following test fails:

pytest -v tests/test_transcribe.py

At the end of this message I have the output I got.

For good measure I execute the first example that is shown in the README.md file. That did not core dump but failed indicating that the Requested float16 compute type was not supported. I have a Quadro P1000. I have the output shown last at the end of this message.

So the example above leads me to believe that access to the GPU is ok. I cannot find detailed specs indicating that float16 is not supported, but that does not seem right.

EDIT: In this link it does seem that fp16 is supported.

So could the failed test be due to low memory? I onky have 4Gb. Xorg is currently using 3MiB.

Any suggestions on how to diagnose and solve this?

TIA,
Hugo

Test failure:

(.venv) hmf@gandalf:/mnt/ssd2/hmf/VSCodeProjects/faster-whisper$ pytest -v tests/test_transcribe.py 
=============================================================== test session starts ===============================================================
platform linux -- Python 3.12.3, pytest-7.4.4, pluggy-1.5.0 -- /mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/bin/python
cachedir: .pytest_cache
rootdir: /mnt/ssd2/hmf/VSCodeProjects/faster-whisper
collected 11 items                                                                                                                                

tests/test_transcribe.py::test_supported_languages PASSED                                                                                   [  9%]
tests/test_transcribe.py::test_transcribe Fatal Python error: Aborted

Thread 0x00007e94ed2006c0 (most recent call first):
  File "/usr/lib/python3.12/threading.py", line 359 in wait
  File "/usr/lib/python3.12/threading.py", line 655 in wait
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/tqdm/_monitor.py", line 60 in run
  File "/usr/lib/python3.12/threading.py", line 1073 in _bootstrap_inner
  File "/usr/lib/python3.12/threading.py", line 1030 in _bootstrap

Thread 0x00007e95da394080 (most recent call first):
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/faster_whisper/transcribe.py", line 1345 in encode
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/faster_whisper/transcribe.py", line 1764 in detect_language
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/faster_whisper/transcribe.py", line 887 in transcribe
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/tests/test_transcribe.py", line 16 in test_transcribe
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/python.py", line 194 in pytest_pyfunc_call
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_callers.py", line 103 in _multicall
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_manager.py", line 120 in _hookexec
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_hooks.py", line 513 in __call__
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/python.py", line 1792 in runtest
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/runner.py", line 169 in pytest_runtest_call
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_callers.py", line 103 in _multicall
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_manager.py", line 120 in _hookexec
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_hooks.py", line 513 in __call__
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/runner.py", line 262 in <lambda>
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/runner.py", line 341 in from_call
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/runner.py", line 261 in call_runtest_hook
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/runner.py", line 222 in call_and_report
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/runner.py", line 133 in runtestprotocol
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/runner.py", line 114 in pytest_runtest_protocol
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_callers.py", line 103 in _multicall
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_manager.py", line 120 in _hookexec
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_hooks.py", line 513 in __call__
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/main.py", line 350 in pytest_runtestloop
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_callers.py", line 103 in _multicall
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_manager.py", line 120 in _hookexec
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_hooks.py", line 513 in __call__
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/main.py", line 325 in _main
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/main.py", line 271 in wrap_session
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/main.py", line 318 in pytest_cmdline_main
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_callers.py", line 103 in _multicall
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_manager.py", line 120 in _hookexec
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/pluggy/_hooks.py", line 513 in __call__
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/config/__init__.py", line 169 in main
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/lib/python3.12/site-packages/_pytest/config/__init__.py", line 192 in console_main
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/bin/pytest", line 8 in <module>

Extension modules: numpy._core._multiarray_umath, numpy.linalg._umath_linalg, av._core, av.logging, av.bytesource, av.buffer, av.audio.format, av.enum, av.error, av.utils, av.option, av.descriptor, av.container.pyio, av.dictionary, av.format, av.stream, av.container.streams, av.sidedata.motionvectors, av.sidedata.sidedata, av.opaque, av.packet, av.container.input, av.container.output, av.container.core, av.codec.context, av.video.format, av.video.reformatter, av.plane, av.video.plane, av.video.frame, av.video.stream, av.codec.codec, av.frame, av.audio.layout, av.audio.plane, av.audio.frame, av.audio.stream, av.filter.pad, av.filter.link, av.filter.context, av.filter.graph, av.filter.filter, av.audio.resampler, av.audio.codeccontext, av.audio.fifo, av.bitstream, av.video.codeccontext, torch._C, torch._C._dynamo.autograd_compiler, torch._C._dynamo.eval_frame, torch._C._dynamo.guards, torch._C._dynamo.utils, torch._C._fft, torch._C._linalg, torch._C._nested, torch._C._nn, torch._C._sparse, torch._C._special, yaml._yaml, charset_normalizer.md, requests.packages.charset_normalizer.md, requests.packages.chardet.md, markupsafe._speedups (total: 63)
Aborted (core dumped)
(.venv) hmf@gandalf:/mnt/ssd2/hmf/VSCodeProjects/faster-whisper$ 

(.venv) hmf@gandalf:/mnt/ssd2/hmf/VSCodeProjects/faster-whisper$ /mnt/ssd2/hmf/VSCodeProjects/faster-whisper/.venv/bin/python /mnt/ssd2/hmf/VSCodeProjects/faster-whisper/exp/example_1.py
preprocessor_config.json: 100%|████████████████████████████████████████████████████████████████████████████████████| 340/340 [00:00<00:00, 619kB/s]
config.json: 100%|████████████████████████████████████████████████████████████████████████████████████████████| 2.39k/2.39k [00:00<00:00, 5.10MB/s]
tokenizer.json: 100%|█████████████████████████████████████████████████████████████████████████████████████████| 2.48M/2.48M [00:01<00:00, 2.01MB/s]
vocabulary.json: 100%|█████████████████████████████████████████████████████████████████████████████████████████| 1.07M/1.07M [00:01<00:00, 872kB/s]
model.bin: 100%|██████████████████████████████████████████████████████████████████████████████████████████████| 3.09G/3.09G [04:32<00:00, 11.3MB/s]
Traceback (most recent call last):
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/exp/example_1.py", line 6, in <module>████████████████████| 3.09G/3.09G [04:32<00:00, 11.3MB/s]
    model = WhisperModel(model_size, device="cuda", compute_type="float16")
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/ssd2/hmf/VSCodeProjects/faster-whisper/faster_whisper/transcribe.py", line 634, in __init__
    self.model = ctranslate2.models.Whisper(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: Requested float16 compute type, but the target device or backend do not support efficient float16 computation.

[heimoshuiyu]

may be #42

[heimoshuiyu]

As far as I know, if your GPU does not support efficient float16 computation (aka tensor core), then its speed is the same as float32.

[hmf]

@heimoshuiyu I will update the libs and report back.

[hmf]

I have realized I have the incorrect version of CuDNN. I have version 8.9.7. The Ubuntu package is up to date so I will have to install directly from NVIDIA. Will report back when done.

[hmf]

The core dump was solved by installing CuDNN 9 (I had version 8.9 installed).
All test executed correctly (had one warning).

I cannot run the example I described above. I get RuntimeError: CUDA failed with error out of memory. This is because I am forced to use float32. I can also use int8 on the GPU. CPU works ok (Quadro P1000 has compute capability 6.1), but is slow as molasses. 8-(.