# Running on vgni027 # Started at Mon 28 Nov 08:09:01 CET 2022 # python3 src/run_speech_recognition_ctc.py --report_to=none --run_name=experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc/ --preprocessing_num_workers=5 --model_name_or_path=facebook/wav2vec2-large-960h-lv60-self --dataset_name=experiments/data/atcosim_uwb_atcc/train --min_duration_in_seconds=0.2 --max_duration_in_seconds=20 --eval_dataset_name=experiments/data/atcosim_uwb_atcc/test --train_split_name=train --output_dir=experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc/ --num_train_epochs=50 --per_device_train_batch_size=24 --per_device_eval_batch_size=12 --gradient_accumulation_steps=1 --learning_rate=1e-4 --weight_decay=0.001 --warmup_steps=1000 --evaluation_strategy=steps --text_column_name=text --audio_column_name=audio --length_column_name=input_length '--chars_to_ignore=, ? . ! \; \: " “ % ‘ ” � — ’ … –' --save_steps=1000 --eval_steps=500 --logging_steps=1000 --layerdrop=0.0 --activation_dropout=0.0 --attention_dropout=0.0 --save_total_limit=1 --feat_proj_dropout=0.05 --mask_time_prob=0.075 --mask_time_length=12 --mask_feature_prob=0.0 --mask_feature_length=12 --gradient_checkpointing --freeze_feature_encoder --fp16 --group_by_length --do_train --do_eval --max_steps 10000 --overwrite_output_dir --freeze_feature_encoder 11/28/2022 08:09:10 - WARNING - __main__ - Process rank: -1, device: cuda:0, n_gpu: 1distributed training: False, 16-bits training: True 11/28/2022 08:09:10 - INFO - __main__ - Training/evaluation parameters TrainingArguments( _n_gpu=1, adafactor=False, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, auto_find_batch_size=False, bf16=False, bf16_full_eval=False, data_seed=None, dataloader_drop_last=False, dataloader_num_workers=0, dataloader_pin_memory=True, ddp_bucket_cap_mb=None, ddp_find_unused_parameters=None, ddp_timeout=1800, debug=[], deepspeed=None, disable_tqdm=False, do_eval=True, do_predict=False, do_train=True, eval_accumulation_steps=None, eval_delay=0, eval_steps=500, evaluation_strategy=steps, fp16=True, fp16_backend=auto, fp16_full_eval=False, fp16_opt_level=O1, fsdp=[], fsdp_min_num_params=0, fsdp_transformer_layer_cls_to_wrap=None, full_determinism=False, gradient_accumulation_steps=1, gradient_checkpointing=True, greater_is_better=None, group_by_length=True, half_precision_backend=auto, hub_model_id=None, hub_private_repo=False, hub_strategy=every_save, hub_token=, ignore_data_skip=False, include_inputs_for_metrics=False, jit_mode_eval=False, label_names=None, label_smoothing_factor=0.0, learning_rate=0.0001, length_column_name=input_length, load_best_model_at_end=False, local_rank=-1, log_level=passive, log_level_replica=passive, log_on_each_node=True, logging_dir=experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc/runs/Nov28_08-09-10_vgni027, logging_first_step=False, logging_nan_inf_filter=True, logging_steps=1000, logging_strategy=steps, lr_scheduler_type=linear, max_grad_norm=1.0, max_steps=10000, metric_for_best_model=None, mp_parameters=, no_cuda=False, num_train_epochs=50.0, optim=adamw_hf, output_dir=experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc/, overwrite_output_dir=True, past_index=-1, per_device_eval_batch_size=12, per_device_train_batch_size=24, prediction_loss_only=False, push_to_hub=False, push_to_hub_model_id=None, push_to_hub_organization=None, push_to_hub_token=, ray_scope=last, remove_unused_columns=True, report_to=[], resume_from_checkpoint=None, run_name=experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc/, save_on_each_node=False, save_steps=1000, save_strategy=steps, save_total_limit=1, seed=42, sharded_ddp=[], skip_memory_metrics=True, tf32=None, torchdynamo=None, tpu_metrics_debug=False, tpu_num_cores=None, use_ipex=False, use_legacy_prediction_loop=False, use_mps_device=False, warmup_ratio=0.0, warmup_steps=1000, weight_decay=0.001, xpu_backend=None, ) 11/28/2022 08:09:10 - WARNING - datasets.builder - Using custom data configuration train-229a51d735ef0a2b Downloading and preparing dataset atc_data_loader/train to /remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/.cache/experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc//train/atc_data_loader/train-229a51d735ef0a2b/0.0.0/f2633cc53c6abe32cddd4152eebde1a4e3c9953e1446e190b8d9a13330cddaa4... Generating train split: 0 examples [00:00, ? examples/s] Generating train split: 1 examples [00:53, 53.46s/ examples] Generating train split: 256 examples [00:53, 6.77 examples/s] Generating train split: 512 examples [00:54, 16.30 examples/s] Generating train split: 768 examples [00:54, 29.69 examples/s] Generating train split: 1024 examples [00:54, 48.26 examples/s] Generating train split: 1280 examples [00:54, 73.60 examples/s] Generating train split: 1536 examples [00:54, 108.14 examples/s] Generating train split: 1792 examples [00:55, 153.96 examples/s] Generating train split: 2048 examples [00:55, 214.00 examples/s] Generating train split: 2304 examples [00:55, 287.75 examples/s] Generating train split: 2560 examples [00:55, 375.64 examples/s] Generating train split: 2816 examples [00:56, 463.83 examples/s] Generating train split: 3072 examples [00:56, 570.48 examples/s] Generating train split: 3328 examples [00:56, 663.67 examples/s] Generating train split: 3584 examples [00:56, 783.88 examples/s] Generating train split: 3840 examples [00:56, 882.22 examples/s] Generating train split: 4096 examples [00:57, 926.29 examples/s] Generating train split: 4352 examples [00:57, 994.41 examples/s] Generating train split: 4608 examples [00:57, 1029.52 examples/s] Generating train split: 4864 examples [00:57, 1022.38 examples/s] Generating train split: 5120 examples [00:58, 1078.20 examples/s] Generating train split: 5376 examples [00:58, 1115.61 examples/s] Generating train split: 5632 examples [00:58, 1118.35 examples/s] Generating train split: 5888 examples [00:58, 1155.38 examples/s] Generating train split: 6144 examples [00:58, 1200.64 examples/s] Generating train split: 6400 examples [00:59, 1166.52 examples/s] Generating train split: 6656 examples [00:59, 1158.81 examples/s] Generating train split: 6912 examples [00:59, 1180.90 examples/s] Generating train split: 7168 examples [00:59, 1154.28 examples/s] Generating train split: 7424 examples [00:59, 1156.77 examples/s] Generating train split: 7680 examples [01:00, 1208.85 examples/s] Generating train split: 8144 examples [01:00, 1786.32 examples/s] Generating train split: 8533 examples [01:00, 2193.75 examples/s] Generating train split: 8960 examples [01:00, 2520.73 examples/s] Generating train split: 9459 examples [01:00, 3083.36 examples/s] Generating train split: 9816 examples [01:00, 3110.19 examples/s] Generating train split: 10240 examples [01:00, 3300.45 examples/s] Generating train split: 10752 examples [01:00, 3521.50 examples/s] Generating train split: 11264 examples [01:01, 3646.93 examples/s] Generating train split: 11776 examples [01:01, 3720.64 examples/s] Generating train split: 12261 examples [01:01, 4003.74 examples/s] Generating train split: 12673 examples [01:01, 3919.00 examples/s] Generating train split: 13073 examples [01:01, 3838.18 examples/s] Generating train split: 13568 examples [01:01, 3925.92 examples/s] Generating train split: 14076 examples [01:01, 4235.56 examples/s] Generating train split: 14525 examples [01:01, 4305.79 examples/s] Generating train split: 14961 examples [01:01, 4162.99 examples/s] Generating train split: 15381 examples [01:02, 4044.01 examples/s] Generating train split: 15872 examples [01:02, 4103.73 examples/s] Generating train split: 16376 examples [01:02, 4361.38 examples/s] Generating train split: 16816 examples [01:02, 4261.34 examples/s] Generating train split: 17245 examples [01:02, 4044.89 examples/s] Generating train split: 17664 examples [01:02, 3868.00 examples/s] Generating train split: 18153 examples [01:02, 4145.65 examples/s] Generating train split: 18573 examples [01:02, 3889.75 examples/s] Dataset atc_data_loader downloaded and prepared to /remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/.cache/experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc//train/atc_data_loader/train-229a51d735ef0a2b/0.0.0/f2633cc53c6abe32cddd4152eebde1a4e3c9953e1446e190b8d9a13330cddaa4. Subsequent calls will reuse this data. 11/28/2022 08:10:24 - WARNING - datasets.builder - Using custom data configuration test-b7cef58c6dbd7d6c Downloading and preparing dataset atc_data_loader/test to /remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/.cache/experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc//test/atc_data_loader/test-b7cef58c6dbd7d6c/0.0.0/f2633cc53c6abe32cddd4152eebde1a4e3c9953e1446e190b8d9a13330cddaa4... Generating test split: 0 examples [00:00, ? examples/s] Generating test split: 1 examples [00:12, 12.03s/ examples] Generating test split: 256 examples [00:12, 29.51 examples/s] Generating test split: 512 examples [00:12, 68.80 examples/s] Generating test split: 768 examples [00:12, 121.23 examples/s] Generating test split: 1024 examples [00:12, 188.38 examples/s] Generating test split: 1280 examples [00:13, 270.78 examples/s] Generating test split: 1536 examples [00:13, 362.18 examples/s] Generating test split: 1792 examples [00:13, 462.16 examples/s] Generating test split: 2048 examples [00:13, 613.65 examples/s] Generating test split: 2480 examples [00:13, 977.92 examples/s] Generating test split: 2816 examples [00:13, 1248.52 examples/s] Generating test split: 3316 examples [00:14, 1800.86 examples/s] Generating test split: 3658 examples [00:14, 2081.76 examples/s] Generating test split: 4096 examples [00:14, 2458.65 examples/s] Generating test split: 4591 examples [00:14, 2996.04 examples/s] loading configuration file config.json from cache at /idiap/temp/jzuluaga/cache/huggingface/models--facebook--wav2vec2-large-960h-lv60-self/snapshots/54074b1c16f4de6a5ad59affb4caa8f2ea03a119/config.json Model config Wav2Vec2Config { "_name_or_path": "facebook/wav2vec2-large-960h-lv60-self", "activation_dropout": 0.1, "adapter_kernel_size": 3, "adapter_stride": 2, "add_adapter": false, "apply_spec_augment": true, "architectures": [ "Wav2Vec2ForCTC" ], "attention_dropout": 0.1, "bos_token_id": 1, "classifier_proj_size": 256, "codevector_dim": 256, "contrastive_logits_temperature": 0.1, "conv_bias": true, "conv_dim": [ 512, 512, 512, 512, 512, 512, 512 ], "conv_kernel": [ 10, 3, 3, 3, 3, 2, 2 ], "conv_stride": [ 5, 2, 2, 2, 2, 2, 2 ], "ctc_loss_reduction": "sum", "ctc_zero_infinity": false, "diversity_loss_weight": 0.1, "do_stable_layer_norm": true, "eos_token_id": 2, "feat_extract_activation": "gelu", "feat_extract_dropout": 0.0, "feat_extract_norm": "layer", "feat_proj_dropout": 0.1, "feat_quantizer_dropout": 0.0, "final_dropout": 0.1, "gradient_checkpointing": false, "hidden_act": "gelu", "hidden_dropout": 0.1, "hidden_dropout_prob": 0.1, "hidden_size": 1024, "initializer_range": 0.02, "intermediate_size": 4096, "layer_norm_eps": 1e-05, "layerdrop": 0.1, "mask_feature_length": 10, "mask_feature_min_masks": 0, "mask_feature_prob": 0.0, "mask_time_length": 10, "mask_time_min_masks": 2, "mask_time_prob": 0.05, "model_type": "wav2vec2", "num_adapter_layers": 3, "num_attention_heads": 16, "num_codevector_groups": 2, "num_codevectors_per_group": 320, "num_conv_pos_embedding_groups": 16, "num_conv_pos_embeddings": 128, "num_feat_extract_layers": 7, "num_hidden_layers": 24, "num_negatives": 100, "output_hidden_size": 1024, "pad_token_id": 0, "proj_codevector_dim": 256, "tdnn_dilation": [ 1, 2, 3, 1, 1 ], "tdnn_dim": [ 512, 512, 512, 512, 1500 ], "tdnn_kernel": [ 5, 3, 3, 1, 1 ], "transformers_version": "4.24.0", "use_weighted_layer_sum": false, "vocab_size": 32, "xvector_output_dim": 512 } Dataset atc_data_loader downloaded and prepared to /remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/.cache/experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc//test/atc_data_loader/test-b7cef58c6dbd7d6c/0.0.0/f2633cc53c6abe32cddd4152eebde1a4e3c9953e1446e190b8d9a13330cddaa4. Subsequent calls will reuse this data. 0%| | 0/1 [00:00 to the vocabulary Adding to the vocabulary Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained. loading configuration file preprocessor_config.json from cache at /idiap/temp/jzuluaga/cache/huggingface/models--facebook--wav2vec2-large-960h-lv60-self/snapshots/54074b1c16f4de6a5ad59affb4caa8f2ea03a119/preprocessor_config.json loading configuration file config.json from cache at /idiap/temp/jzuluaga/cache/huggingface/models--facebook--wav2vec2-large-960h-lv60-self/snapshots/54074b1c16f4de6a5ad59affb4caa8f2ea03a119/config.json Model config Wav2Vec2Config { "_name_or_path": "facebook/wav2vec2-large-960h-lv60-self", "activation_dropout": 0.1, "adapter_kernel_size": 3, "adapter_stride": 2, "add_adapter": false, "apply_spec_augment": true, "architectures": [ "Wav2Vec2ForCTC" ], "attention_dropout": 0.1, "bos_token_id": 1, "classifier_proj_size": 256, "codevector_dim": 256, "contrastive_logits_temperature": 0.1, "conv_bias": true, "conv_dim": [ 512, 512, 512, 512, 512, 512, 512 ], "conv_kernel": [ 10, 3, 3, 3, 3, 2, 2 ], "conv_stride": [ 5, 2, 2, 2, 2, 2, 2 ], "ctc_loss_reduction": "sum", "ctc_zero_infinity": false, "diversity_loss_weight": 0.1, "do_stable_layer_norm": true, "eos_token_id": 2, "feat_extract_activation": "gelu", "feat_extract_dropout": 0.0, "feat_extract_norm": "layer", "feat_proj_dropout": 0.1, "feat_quantizer_dropout": 0.0, "final_dropout": 0.1, "gradient_checkpointing": false, "hidden_act": "gelu", "hidden_dropout": 0.1, "hidden_dropout_prob": 0.1, "hidden_size": 1024, "initializer_range": 0.02, "intermediate_size": 4096, "layer_norm_eps": 1e-05, "layerdrop": 0.1, "mask_feature_length": 10, "mask_feature_min_masks": 0, "mask_feature_prob": 0.0, "mask_time_length": 10, "mask_time_min_masks": 2, "mask_time_prob": 0.05, "model_type": "wav2vec2", "num_adapter_layers": 3, "num_attention_heads": 16, "num_codevector_groups": 2, "num_codevectors_per_group": 320, "num_conv_pos_embedding_groups": 16, "num_conv_pos_embeddings": 128, "num_feat_extract_layers": 7, "num_hidden_layers": 24, "num_negatives": 100, "output_hidden_size": 1024, "pad_token_id": 0, "proj_codevector_dim": 256, "tdnn_dilation": [ 1, 2, 3, 1, 1 ], "tdnn_dim": [ 512, 512, 512, 512, 1500 ], "tdnn_kernel": [ 5, 3, 3, 1, 1 ], "transformers_version": "4.24.0", "use_weighted_layer_sum": false, "vocab_size": 32, "xvector_output_dim": 512 } Feature extractor Wav2Vec2FeatureExtractor { "do_normalize": true, "feature_extractor_type": "Wav2Vec2FeatureExtractor", "feature_size": 1, "padding_side": "right", "padding_value": 0.0, "return_attention_mask": true, "sampling_rate": 16000 } loading weights file pytorch_model.bin from cache at /idiap/temp/jzuluaga/cache/huggingface/models--facebook--wav2vec2-large-960h-lv60-self/snapshots/54074b1c16f4de6a5ad59affb4caa8f2ea03a119/pytorch_model.bin All model checkpoint weights were used when initializing Wav2Vec2ForCTC. Some weights of Wav2Vec2ForCTC were not initialized from the model checkpoint at facebook/wav2vec2-large-960h-lv60-self and are newly initialized: ['wav2vec2.masked_spec_embed'] You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. Some weights of Wav2Vec2ForCTC were not initialized from the model checkpoint at facebook/wav2vec2-large-960h-lv60-self and are newly initialized because the shapes did not match: - lm_head.weight: found shape torch.Size([32, 1024]) in the checkpoint and torch.Size([31, 1024]) in the model instantiated - lm_head.bias: found shape torch.Size([32]) in the checkpoint and torch.Size([31]) in the model instantiated You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. preprocess datasets #0: 0%| | 0/3786 [00:00 to the vocabulary Adding to the vocabulary max_steps is given, it will override any value given in num_train_epochs Using cuda_amp half precision backend The following columns in the training set don't have a corresponding argument in `Wav2Vec2ForCTC.forward` and have been ignored: input_length. If input_length are not expected by `Wav2Vec2ForCTC.forward`, you can safely ignore this message. /idiap/user/jzuluaga/miniconda3/envs/w2v2/lib/python3.10/site-packages/transformers/optimization.py:306: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use the PyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning warnings.warn( ***** Running training ***** Num examples = 18925 Num Epochs = 13 Instantaneous batch size per device = 24 Total train batch size (w. parallel, distributed & accumulation) = 24 Gradient Accumulation steps = 1 Total optimization steps = 10000 Number of trainable parameters = 311260319 0%| | 0/10000 [00:00