18liumin
commented
3 months ago 模型的输入维度是[1, 30, 43]
Open 18liumin opened 3 months ago
18liumin
commented
3 months ago 模型的输入维度是[1, 30, 43]
Donaldkidd
commented
3 months ago 建议你把单独的bn换成1*1卷积试试
[Warning] Isolated BatchNormalization(/encoder/fea_attention_layers.0/fea_attention_layers.0.0/bn/BatchNormalization) was detected, PPQ will replace it to 11 Convolution(2D). Traceback (most recent call last): File "/home/l00576713/data/megaton-LM/package/temp/profile/test2.py", line 24, in
ir = quantize_torch_model(
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/core/defs.py", line 54, in _wrapper
return func( args, *kwargs)
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/api/interface.py", line 343, in quantize_torch_model
return quantize_onnx_model(onnx_import_file=onnx_export_file,
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/core/defs.py", line 54, in _wrapper
return func(args, **kwargs)
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/api/interface.py", line 251, in quantize_onnx_model
ppq_ir = load_onnx_graph(onnx_import_file=onnx_import_file)
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/api/interface.py", line 50, in load_onnx_graph
return format_graph(graph=ppq_ir)
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/api/interface.py", line 627, in format_graph
formatter(GraphCommand(GraphCommandType.REPLACE_BATCHNORM_TO_CONV))
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/IR/processer.py", line 130, in call
return self.process(command)
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/IR/morph.py", line 28, in process
return self.replace_batchnorm_to_conv()
File "/home/l00576713/software/anaconda3/envs/megaton/lib/python3.8/site-packages/ppq/IR/morph.py", line 89, in replace_batchnorm_to_conv
assert len(op.parameters) == 4, "BatchNorm should have 4 parameters, namely alpha, beta, mean, var"
AssertionError: BatchNorm should have 4 parameters, namely alpha, beta, mean, var