[MusicGen] Add sampling rate to config #26136
Conversation
|
|
||
| forward_params = {"max_new_tokens": 10} | ||
| # test single sample | ||
| outputs = music_generator("this", forward_params=forward_params) |
There was a problem hiding this comment.
Can only use very limited prompts for the tiny model since it has a vocab size of 99 😅
|
|
||
| forward_params = { | ||
| "do_sample": False, | ||
| "max_new_tokens": 250, | ||
| } | ||
|
|
||
| outputs = speech_generator("This is a test", forward_params=forward_params) | ||
| # musicgen sampling_rate is not straightforward to get |
There was a problem hiding this comment.
Fixed by this PR! The remainder of the changes are just small refactors to the pipeline tests
|
The documentation is not available anymore as the PR was closed or merged. |
There was a problem hiding this comment.
I just have a small question regarding testing the tiny musicgen, but otherwise thanks for taking care of this @sanchit-gandhi and for the clearer renaming !
| @require_torch | ||
| def test_tiny_musicgen_pt(self): | ||
| music_generator = pipeline( | ||
| task="text-to-audio", | ||
| model="hf-internal-testing/tiny-random-MusicgenForConditionalGeneration", | ||
| framework="pt", | ||
| ) |
There was a problem hiding this comment.
I might be wrong, but models that are in hf-internal-testing should be tested with run_pipeline_test called by PipelineTesterMixin, right ?
Maybe @ydshieh can answer this question ?
There was a problem hiding this comment.
Good spot - MusicGen currently isn't tested by the pipeline, but I'll fix this in a follow-up PR: #26146
Co-authored-by: Arthur <[email protected]>
* [MusicGen] Add sampling rate to config * remove tiny * make property * Update tests/pipelines/test_pipelines_text_to_audio.py Co-authored-by: Arthur <[email protected]> * style --------- Co-authored-by: Arthur <[email protected]>
* [MusicGen] Add sampling rate to config * remove tiny * make property * Update tests/pipelines/test_pipelines_text_to_audio.py Co-authored-by: Arthur <[email protected]> * style --------- Co-authored-by: Arthur <[email protected]>
* [MusicGen] Add sampling rate to config * remove tiny * make property * Update tests/pipelines/test_pipelines_text_to_audio.py Co-authored-by: Arthur <[email protected]> * style --------- Co-authored-by: Arthur <[email protected]>
What does this PR do?
Adds sampling rate attribute to the parent model config. We get the sampling rate from the EnCodec sub-model config. Adding this attribute brings consistency with the other TTA models in the lib, where the output sampling rate is accessible directly through the top-level config (huggingface/audio-transformers-course#140 (comment))