W0423 17:35:18.793000 5422 torch/distributed/run.py:792] W0423 17:35:18.793000 5422 torch/distributed/run.py:792] ***************************************** W0423 17:35:18.793000 5422 torch/distributed/run.py:792] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. W0423 17:35:18.793000 5422 torch/distributed/run.py:792] ***************************************** /root/workdir/.venv/lib/python3.10/site-packages/transformers/training_args.py:2085: FutureWarning: `--push_to_hub_organization` is deprecated and will be removed in version 5 of 🤗 Transformers. Use `--hub_model_id` instead and pass the full repo name to this argument (in this case CLEAR-Global/w2v-bert-2.0-hausa_naijavoices_100h). warnings.warn( 04/23/2025 17:35:24 - WARNING - __main__ - Process rank: 0, device: cuda:0, n_gpu: 1, distributed training: True, 16-bits training: True 04/23/2025 17:35:24 - INFO - __main__ - Training/evaluation parameters TrainingArguments( _n_gpu=1, accelerator_config={'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None, 'use_configured_state': False}, adafactor=False, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, auto_find_batch_size=False, average_tokens_across_devices=False, batch_eval_metrics=False, bf16=False, bf16_full_eval=False, data_seed=None, dataloader_drop_last=False, dataloader_num_workers=44, dataloader_persistent_workers=False, dataloader_pin_memory=True, dataloader_prefetch_factor=None, ddp_backend=None, ddp_broadcast_buffers=None, ddp_bucket_cap_mb=None, ddp_find_unused_parameters=None, ddp_timeout=1800, debug=[], deepspeed=None, disable_tqdm=False, dispatch_batches=None, do_eval=True, do_predict=False, do_train=True, eval_accumulation_steps=None, eval_delay=0, eval_do_concat_batches=True, eval_on_start=False, eval_steps=1000, eval_strategy=steps, eval_use_gather_object=False, evaluation_strategy=None, fp16=True, fp16_backend=auto, fp16_full_eval=False, fp16_opt_level=O1, fsdp=[], fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, fsdp_min_num_params=0, fsdp_transformer_layer_cls_to_wrap=None, full_determinism=False, gradient_accumulation_steps=1, gradient_checkpointing=True, gradient_checkpointing_kwargs=None, greater_is_better=False, group_by_length=True, half_precision_backend=auto, hub_always_push=False, hub_model_id=CLEAR-Global/w2v-bert-2.0-hausa_naijavoices_100h, hub_private_repo=None, hub_strategy=checkpoint, hub_token=, ignore_data_skip=False, include_for_metrics=[], include_inputs_for_metrics=False, include_num_input_tokens_seen=False, include_tokens_per_second=False, jit_mode_eval=False, label_names=None, label_smoothing_factor=0.0, learning_rate=3e-05, length_column_name=input_length, load_best_model_at_end=True, local_rank=0, log_level=passive, log_level_replica=warning, log_on_each_node=True, logging_dir=./w2v-bert-2.0-hausa_naijavoices_100h/runs/Apr23_17-35-24_synvoices-hausa-2, logging_first_step=False, logging_nan_inf_filter=True, logging_steps=1.0, logging_strategy=steps, lr_scheduler_kwargs={}, lr_scheduler_type=linear, max_grad_norm=1.0, max_steps=-1, metric_for_best_model=loss, mp_parameters=, neftune_noise_alpha=None, no_cuda=False, num_train_epochs=250.0, optim=adamw_torch, optim_args=None, optim_target_modules=None, output_dir=./w2v-bert-2.0-hausa_naijavoices_100h, overwrite_output_dir=False, past_index=-1, per_device_eval_batch_size=160, per_device_train_batch_size=160, prediction_loss_only=False, push_to_hub=True, push_to_hub_model_id=None, push_to_hub_organization=CLEAR-Global, push_to_hub_token=, ray_scope=last, remove_unused_columns=True, report_to=['tensorboard'], restore_callback_states_from_checkpoint=False, resume_from_checkpoint=None, run_name=./w2v-bert-2.0-hausa_naijavoices_100h, save_on_each_node=False, save_only_model=False, save_safetensors=True, save_steps=1000, save_strategy=steps, save_total_limit=1, seed=42, skip_memory_metrics=True, split_batches=None, tf32=None, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, torch_empty_cache_steps=None, torchdynamo=None, tpu_metrics_debug=False, tpu_num_cores=None, use_cpu=False, use_ipex=False, use_legacy_prediction_loop=False, use_liger_kernel=False, use_mps_device=False, warmup_ratio=0.1, warmup_steps=0, weight_decay=0.0, ) /root/workdir/.venv/lib/python3.10/site-packages/transformers/training_args.py:2085: FutureWarning: `--push_to_hub_organization` is deprecated and will be removed in version 5 of 🤗 Transformers. Use `--hub_model_id` instead and pass the full repo name to this argument (in this case CLEAR-Global/w2v-bert-2.0-hausa_naijavoices_100h). warnings.warn( 04/23/2025 17:35:24 - WARNING - __main__ - Process rank: 1, device: cuda:1, n_gpu: 1, distributed training: True, 16-bits training: True Downloading data: 0%| | 0/24 [00:00', 'eos_token': '', 'unk_token': '[UNK]', 'pad_token': '[PAD]'}, clean_up_tokenization_spaces=False, added_tokens_decoder={ 85: AddedToken("[UNK]", rstrip=True, lstrip=True, single_word=False, normalized=False, special=False), 86: AddedToken("[PAD]", rstrip=True, lstrip=True, single_word=False, normalized=False, special=False), 87: AddedToken("", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True), 88: AddedToken("", rstrip=False, lstrip=False, single_word=False, normalized=False, special=True), } ) { "processor_class": "Wav2Vec2BertProcessor" } Using auto half precision backend 04/23/2025 17:39:26 - INFO - __main__ - Fine-tuning model from scratch The following columns in the training set don't have a corresponding argument in `Wav2Vec2BertForCTC.forward` and have been ignored: input_length. If input_length are not expected by `Wav2Vec2BertForCTC.forward`, you can safely ignore this message. ***** Running training ***** Num examples = 114,548 Num Epochs = 250 Instantaneous batch size per device = 160 Total train batch size (w. parallel, distributed & accumulation) = 320 Gradient Accumulation steps = 1 Total optimization steps = 89,500 Number of trainable parameters = 605,768,601 0%| | 0/89500 [00:00