anyway to run it on amd cards like rx6800 ?
audio loading part is mostly cuda if i am not mistaking
i try to make it work by changing it to cpu base and i got this error
`Exception has occurred: RuntimeError
The size of tensor a (50) must match the size of tensor b (20) at non-singleton dimension 1
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/src/modeling_moss_music.py", line 148, in _encode_chunk_batch
x,
attention_mask,
layer_head_mask=None,
output_attentions=False,
)[0]
capture_idx = self._deepstack_capture_map.get(layer_idx)
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/src/modeling_moss_music.py", line 257, in forward
out, deepstack_outs = self._encode_chunk_batch(feat_chunk, len_chunk)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/src/modeling_moss_music.py", line 385, in get_audio_features
input_features=input_features,
feature_lens=feature_lens,
output_deepstack_hidden_states=True,
)
deepstack = (
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/src/modeling_moss_music.py", line 488, in forward
audio_embeds, deepstack = self.get_audio_features(audio_data, audio_data_seqlens)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/infer.py", line 44, in main
**inputs,
max_new_tokens=1024,
do_sample=True,
num_beams=1,
temperature=TEMPERATURE,
top_p=TOP_P,
top_k=TOP_K,
use_cache=True,
)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/infer.py", line 66, in
main()
RuntimeError: The size of tensor a (50) must match the size of tensor b (20) at non-singleton dimension 1`
anyway to run it on amd cards like rx6800 ?
audio loading part is mostly cuda if i am not mistaking
i try to make it work by changing it to cpu base and i got this error
`Exception has occurred: RuntimeError
The size of tensor a (50) must match the size of tensor b (20) at non-singleton dimension 1
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/src/modeling_moss_music.py", line 148, in _encode_chunk_batch
x,
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/src/modeling_moss_music.py", line 257, in forward
out, deepstack_outs = self._encode_chunk_batch(feat_chunk, len_chunk)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/src/modeling_moss_music.py", line 385, in get_audio_features
input_features=input_features,
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/src/modeling_moss_music.py", line 488, in forward
audio_embeds, deepstack = self.get_audio_features(audio_data, audio_data_seqlens)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/infer.py", line 44, in main
**inputs,
File "/home/ahmad/Desktop/python12/New Folder/MOSS-Music/infer.py", line 66, in
main()
RuntimeError: The size of tensor a (50) must match the size of tensor b (20) at non-singleton dimension 1`