state = {
'model': {
"name": crnn.__class__.__name__,
'args': '',
"kwargs": crnn_kwargs,
'state_dict': crnn.state_dict()
},
'optimizer': {
"name": optimizer.__class__.__name__,
'args': '',
"kwargs": optim_kwargs,
'state_dict': optimizer.state_dict()
},
"pooling_time_ratio": pooling_time_ratio,
'scaler': scaler.state_dict(),
"many_hot_encoder": many_hot_encoder.state_dict()
}
save_best_cb = SaveBest("sup")
# Eval 2018
eval_2018_df = dataset.initialize_and_get_df(cfg.eval2018,
reduced_number_of_data)
eval_2018 = DataLoadDf(eval_2018_df,
dataset.get_feature_file,
many_hot_encoder.encode_strong_df,
transform=transforms_valid)
[crnn] = to_cuda_if_available([crnn])
for epoch in range(cfg.n_epoch):
crnn = crnn.train()
print("min classes examples: " + str(min_length))
number = min_length * 10
print('number valid: ' + str(number))
test_triplets = DataLoadDfTripletLabeledExhaustif(test_df_fr,
encode_function_label,
transform=Compose(trans_fr_sc_embed),
number=min(cfg.number_test, len(test_df_fr.dropna())))
# #########
# End of DATA
# ########
if resume_training is None:
state.update({
'scaler': scaler.state_dict(),
'many_hot_encoder': many_hot_encoder.state_dict()
})
model_directory, log_directory = get_dirs("pretrained" + "_bs_" + str(batch_size) + "adam")
if frames_in_sec is not None:
fr = frames_in_sec
elif segment:
fr = "seg"
else:
fr = "unknown"
params_name = {
"early_stopping": cfg.early_stopping,
"conv_dropout": cfg.conv_dropout,
"frames": fr,
}
if not args.segment:
trans_emb.append(Unsqueeze(0))
train_set_emb = DataLoadDf(train_weak_df,
many_hot_encoder.encode_weak,
transform=Compose(trans_emb))
valid_set_val = DataLoadDf(valid_weak_df,
many_hot_encoder.encode_weak,
transform=Compose(trans_emb))
test_set_val = DataLoadDf(test_df,
many_hot_encoder.encode_weak,
transform=Compose(trans_emb))
emb_state = {
"scaler": scaler.state_dict(),
"many_hot_encoder": many_hot_encoder.state_dict()
}
emb_model, emb_state = get_model(emb_state, args)
emb_model = to_cuda_if_available(emb_model)
# Classif_model
if args.segment:
X, y = train_set[0]
else:
X, y = next(iter(train_load))
X = to_cuda_if_available(X)
emb = emb_model(X)
LOG.info("shape input CNN: x {}, y {}".format(X.shape, y.shape))
LOG.info("shape after CNN: {}".format(emb.shape))
if args.n_layers_classif == 2:
dimensions = [32, 16]