An open source AutoML toolkit for automate machine learning lifecycle, including feature engineering, neural architecture search, model compression and hyper-parameter tuning.
but the pruning gets killed because it rans out of memory.here is part of my output.
.
.0:492: 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!
ne_15 = size_32 != (mul_31, getitem_27, 80); size_32 = mul_31 = None
.
.
[2023-09-04 13:04:03] Propagate variables for call_function: getitem_121
[2023-09-04 13:04:03] Propagate variables for call_module: wav2vec2_encoder_layers_38_attention_q_proj
[2023-09-04 13:04:03] Propagate variables for call_function: mul_152
[2023-09-04 13:04:03] Propagate variables for call_module: wav2vec2_encoder_layers_38_attention_k_proj
[2023-09-04 13:04:03] Propagate variables for call_method: view_190
[2023-09-04 13:04:03] Propagate variables for call_method: transpose_207
[2023-09-04 13:04:03] Propagate variables for call_method: contiguous_114
[2023-09-04 13:04:03] Propagate variables for call_module: wav2vec2_encoder_layers_38_attention_v_proj
[2023-09-04 13:04:03] Propagate variables for call_method: view_191
[2023-09-04 13:04:03] Propagate variables for call_method: transpose_208
[2023-09-04 13:04:03] Propagate variables for call_method: contiguous_115
[2023-09-04 13:04:03] Propagate variables for call_function: mul_153
[2023-09-04 13:04:03] Propagate variables for call_method: view_192
[2023-09-04 13:04:03] Propagate variables for call_method: transpose_209
[2023-09-04 13:04:03] Propagate variables for call_method: contiguous_116
[2023-09-04 13:04:03] Propagate variables for call_method: view_193
[2023-09-04 13:04:03] Propagate variables for call_method: reshape_114
[2023-09-04 13:04:04] Propagate variables for call_method: reshape_115
Killed
my model is 3.86 G but when pruning it uses around 100 G ram.
whats the problem?do you know the reason?
any help is really appreciated!
@J-shang hi,thanks for your great work. i am trying to prune wav2vec2 model from huggingface and i am using the code below:
but the pruning gets killed because it rans out of memory.here is part of my output.
.