Upload 11 files

config.json

@@ -0,0 +1,116 @@
+{
+  "_name_or_path": "Edresson/wav2vec2-large-xlsr-coraa-portuguese",
+  "activation_dropout": 0.0,
+  "adapter_attn_dim": null,
+  "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": 768,
+  "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": "mean",
+  "ctc_zero_infinity": true,
+  "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": 0.0,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.0,
+  "mask_channel_length": 10,
+  "mask_channel_min_space": 1,
+  "mask_channel_other": 0.0,
+  "mask_channel_prob": 0.0,
+  "mask_channel_selection": "static",
+  "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_min_space": 1,
+  "mask_time_other": 0.0,
+  "mask_time_prob": 0.05,
+  "mask_time_selection": "static",
+  "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": 768,
+  "tdnn_dilation": [
+    1,
+    2,
+    3,
+    1,
+    1
+  ],
+  "tdnn_dim": [
+    512,
+    512,
+    512,
+    512,
+    1500
+  ],
+  "tdnn_kernel": [
+    5,
+    3,
+    3,
+    1,
+    1
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.39.3",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 45,
+  "xvector_output_dim": 512
+}

config_train.json

@@ -0,0 +1,94 @@
+{
+    "run_name": "Wav2Vec-fine-tuning-NURC-SP",
+    "run_description": "Fine tuning NURC-SP",
+    "seed": 42,
+    // AUDIO PARAMS
+    "sampling_rate": 16000,
+
+    // VOCABULARY PARAMETERS
+    "vocab":{
+        "vocab_path": "example/vocab_example.json", // generic vocab for Portuguese
+        "blank": "<pad>", // blank token for padding
+        "silence": "|", // token between words
+        "unk": "<unk>" // unk token
+    },
+
+    // TRAINING
+    "batch_size": 8,       // Batch size for training.
+    "mixed_precision": true,     // level of optimization with NVIDIA's apex feature for automatic mixed FP16/FP32 precision (AMP), NOTE: currently only O1 is supported, and use "O1" to activate.
+    "early_stop_epochs": 10, // If 0 disabled else Number of epochs for stop training with validation loss dont decrease 
+    "preprocess_dataset": false, // if true, the dataset will be pre-processed and saved in disk, otherwise the audio files will be loaded in each step. Preprocessing makes training faster, but requires much more disk space.
+
+    // OPTIMIZER
+    "epochs": 50,                // total number of epochs to train.
+    "lr": 0.00003,                  // Initial learning rate.
+    "gradient_accumulation_steps": 24, 
+
+    // LOGGING
+    "logging_steps": 100,    // Number of steps to plot.
+    "load_best_model_at_end": true,
+    "save_total_limit": 3,
+    "warmup_ratio": 0.05, // 0 disable Ratio of total training steps used for a linear warmup from 0 to learning_rate
+    "warmup_steps": 0, // 0 disable  Number of steps used for a linear warmup from 0 to learning_rate
+
+    // DATA LOADING
+    "num_loader_workers": 8,        // number of training data loader processes. Don't set it too big. 4-8 are goo
+    
+    // MODEL
+    "freeze_feature_encoder": true, // Whether to freeze the feature encoder layers of the model.
+    "attention_dropout": 0.1, // The dropout ratio for the attention probabilities.
+    "activation_dropout": 0.1, // The dropout ratio for activations inside the fully connected layer.
+    "hidden_dropout": 0.1, // The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
+    "feat_proj_dropout": 0.1, // The dropout probabilitiy for all 1D convolutional layers in feature encoder.
+    "mask_time_prob": 0.05, //  Propability of each feature vector along the time axis to be chosen as the start of the vector span to be masked.
+    "layerdrop": 0.0,  // The LayerDrop probability. 
+    "gradient_checkpointing": true,  // If True, use gradient checkpointing to save memory at the expense of slower backward pass.
+
+
+    // PATHS
+    "output_path": "../checkpoints/Wav2Vec/NURC-SP/final-version/train/",
+    // CACHE
+    "dataset_cache": "../datasets/",
+
+    // DATASETS
+    "datasets":{
+        "files_path": "../datasets/NURC-SP/audios/", // relative path for audios It's will be join with the audio path CSV
+        "train":
+            [
+               // this dicts is pass directly for the load dataset see the documentation: https://huggingface.co/docs/datasets/package_reference/loading_methods.html#datasets.load_dataset 
+            {
+                "name": "csv",
+                "path": "csv",
+                "data_files": ["../datasets/NURC-SP/corpus_2_train.csv"], // csv files
+                "text_column": "text",
+                "path_column": "file_path"
+            }
+            ]
+        ,
+        "devel":
+            [ 
+                {
+                    "name": "csv",
+                    "path": "csv",
+                    "data_files": ["../datasets/NURC-SP/corpus_2_dev.csv"], // csv files
+                    "text_column": "text",
+                    "path_column": "file_path"
+                }
+        
+            ]
+    }//,
+    // used only for test 
+    // "KenLM":{
+    // "kenlm_model_path": "../../kenLM/binaries/subtitle/4-gram/lm.binary", // Path for KenLM model
+    // "lexicon_path": "example/lexicon.lst", // file with all words for limit the decoder search
+    // "beam": 2048, 
+    // "nbest": 1,
+    // "beam_threshold": 25, 
+    // "lm_weight": 1, 
+    // "word_score": -1,
+    // "sil_weight": 0
+    // }
+
+
+}
+

eval_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 16.0,
+    "eval_loss": 0.47797250747680664,
+    "eval_runtime": 91.602,
+    "eval_samples": 3135,
+    "eval_samples_per_second": 34.224,
+    "eval_steps_per_second": 4.279,
+    "eval_wer": 0.23347451217667523
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:493dd95b8bd989f4026512f5285fed16a1d130d9b822c4d79eff5320257f3f2f
+size 1261991980

preprocessor_config.json

@@ -0,0 +1,10 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "processor_class": "Wav2Vec2Processor",
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "<s>",
+  "eos_token": "</s>",
+  "pad_token": "<pad>",
+  "unk_token": "<unk>"
+}

tokenizer_config.json

@@ -0,0 +1,48 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<pad>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "2": {
+      "content": "<unk>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "43": {
+      "content": "<s>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "44": {
+      "content": "</s>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "do_lower_case": false,
+  "eos_token": "</s>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "processor_class": "Wav2Vec2Processor",
+  "replace_word_delimiter_char": " ",
+  "target_lang": null,
+  "tokenizer_class": "Wav2Vec2CTCTokenizer",
+  "unk_token": "<unk>",
+  "word_delimiter_token": "|"
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 16.0,
+    "train_loss": 103989713142.67763,
+    "train_runtime": 155022.8715,
+    "train_samples": 166630,
+    "train_samples_per_second": 53.744,
+    "train_steps_per_second": 0.28
+}

trainer_state.json

@@ -0,0 +1,1147 @@
+{
+  "best_metric": 0.47799625992774963,
+  "best_model_checkpoint": "../checkpoints/Wav2Vec/NURC-SP/final-version/train/checkpoint-5207",
+  "epoch": 16.0,
+  "eval_steps": 500,
+  "global_step": 13886,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "grad_norm": 0.48270344734191895,
+      "learning_rate": 1.3837638376383764e-08,
+      "loss": 0.4969,
+      "step": 1
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": NaN,
+      "learning_rate": 1.3699261992619927e-06,
+      "loss": 1.0935,
+      "step": 100
+    },
+    {
+      "epoch": 0.23,
+      "grad_norm": 10.196231842041016,
+      "learning_rate": 2.753690036900369e-06,
+      "loss": 1.0424,
+      "step": 200
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 7.084147930145264,
+      "learning_rate": 4.137453874538745e-06,
+      "loss": 0.9727,
+      "step": 300
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 4.268230438232422,
+      "learning_rate": 5.521217712177122e-06,
+      "loss": 0.9538,
+      "step": 400
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 7.1523942947387695,
+      "learning_rate": 6.904981549815498e-06,
+      "loss": 0.941,
+      "step": 500
+    },
+    {
+      "epoch": 0.69,
+      "grad_norm": 7.721753120422363,
+      "learning_rate": 8.288745387453875e-06,
+      "loss": 0.9294,
+      "step": 600
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 6.676694393157959,
+      "learning_rate": 9.67250922509225e-06,
+      "loss": 0.9104,
+      "step": 700
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 13.917472839355469,
+      "learning_rate": 1.1056273062730627e-05,
+      "loss": 0.9171,
+      "step": 800
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 0.5722900629043579,
+      "eval_runtime": 102.9565,
+      "eval_samples_per_second": 30.45,
+      "eval_steps_per_second": 3.807,
+      "eval_wer": 0.25805475722281046,
+      "step": 867
+    },
+    {
+      "epoch": 1.04,
+      "grad_norm": 1.1242845058441162,
+      "learning_rate": 1.2440036900369004e-05,
+      "loss": 0.8226,
+      "step": 900
+    },
+    {
+      "epoch": 1.15,
+      "grad_norm": 1.4485478401184082,
+      "learning_rate": 1.382380073800738e-05,
+      "loss": 0.8966,
+      "step": 1000
+    },
+    {
+      "epoch": 1.27,
+      "grad_norm": 1.5878037214279175,
+      "learning_rate": 1.5207564575645757e-05,
+      "loss": 0.881,
+      "step": 1100
+    },
+    {
+      "epoch": 1.38,
+      "grad_norm": 0.9301995635032654,
+      "learning_rate": 1.6591328413284134e-05,
+      "loss": 0.8691,
+      "step": 1200
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 1.2811486721038818,
+      "learning_rate": 1.797509225092251e-05,
+      "loss": 0.8756,
+      "step": 1300
+    },
+    {
+      "epoch": 1.61,
+      "grad_norm": 1.5193356275558472,
+      "learning_rate": 1.9358856088560884e-05,
+      "loss": 0.8687,
+      "step": 1400
+    },
+    {
+      "epoch": 1.73,
+      "grad_norm": 1.1962685585021973,
+      "learning_rate": 2.072878228782288e-05,
+      "loss": 0.8588,
+      "step": 1500
+    },
+    {
+      "epoch": 1.84,
+      "grad_norm": 1.906406044960022,
+      "learning_rate": 2.2112546125461254e-05,
+      "loss": 0.8614,
+      "step": 1600
+    },
+    {
+      "epoch": 1.96,
+      "grad_norm": 2.207412004470825,
+      "learning_rate": 2.349630996309963e-05,
+      "loss": 0.8574,
+      "step": 1700
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 0.5288712978363037,
+      "eval_runtime": 107.9908,
+      "eval_samples_per_second": 29.03,
+      "eval_steps_per_second": 3.63,
+      "eval_wer": 0.254853022739878,
+      "step": 1735
+    },
+    {
+      "epoch": 2.07,
+      "grad_norm": 1.1061471700668335,
+      "learning_rate": 2.4880073800738008e-05,
+      "loss": 0.8829,
+      "step": 1800
+    },
+    {
+      "epoch": 2.19,
+      "grad_norm": 0.760769784450531,
+      "learning_rate": 2.6263837638376385e-05,
+      "loss": 0.8459,
+      "step": 1900
+    },
+    {
+      "epoch": 2.3,
+      "grad_norm": 1.408493161201477,
+      "learning_rate": 2.764760147601476e-05,
+      "loss": 0.8378,
+      "step": 2000
+    },
+    {
+      "epoch": 2.42,
+      "grad_norm": 0.9650698304176331,
+      "learning_rate": 2.9031365313653138e-05,
+      "loss": 0.8415,
+      "step": 2100
+    },
+    {
+      "epoch": 2.53,
+      "grad_norm": 0.9561858177185059,
+      "learning_rate": 2.9978145791850807e-05,
+      "loss": 0.8364,
+      "step": 2200
+    },
+    {
+      "epoch": 2.65,
+      "grad_norm": 0.7619096040725708,
+      "learning_rate": 2.9905298431353505e-05,
+      "loss": 0.8364,
+      "step": 2300
+    },
+    {
+      "epoch": 2.77,
+      "grad_norm": 1.0944970846176147,
+      "learning_rate": 2.9833179544461174e-05,
+      "loss": 0.8256,
+      "step": 2400
+    },
+    {
+      "epoch": 2.88,
+      "grad_norm": 0.7943573594093323,
+      "learning_rate": 2.9760332183963868e-05,
+      "loss": 0.8175,
+      "step": 2500
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 2.035665988922119,
+      "learning_rate": 2.9688213297071538e-05,
+      "loss": 0.8533,
+      "step": 2600
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 0.5230820178985596,
+      "eval_runtime": 99.7653,
+      "eval_samples_per_second": 31.424,
+      "eval_steps_per_second": 3.929,
+      "eval_wer": 0.2511722886098926,
+      "step": 2603
+    },
+    {
+      "epoch": 3.11,
+      "grad_norm": 1.9500775337219238,
+      "learning_rate": 2.9615365936574232e-05,
+      "loss": 0.8022,
+      "step": 2700
+    },
+    {
+      "epoch": 3.23,
+      "grad_norm": 2.2833547592163086,
+      "learning_rate": 2.954251857607693e-05,
+      "loss": 0.8143,
+      "step": 2800
+    },
+    {
+      "epoch": 3.34,
+      "grad_norm": 1.7430369853973389,
+      "learning_rate": 2.9469671215579623e-05,
+      "loss": 0.8134,
+      "step": 2900
+    },
+    {
+      "epoch": 3.46,
+      "grad_norm": 1.7504712343215942,
+      "learning_rate": 2.9396823855082317e-05,
+      "loss": 0.8026,
+      "step": 3000
+    },
+    {
+      "epoch": 3.57,
+      "grad_norm": 1.6993814706802368,
+      "learning_rate": 2.9323976494585015e-05,
+      "loss": 0.8098,
+      "step": 3100
+    },
+    {
+      "epoch": 3.69,
+      "grad_norm": 3.484405279159546,
+      "learning_rate": 2.925112913408771e-05,
+      "loss": 0.812,
+      "step": 3200
+    },
+    {
+      "epoch": 3.8,
+      "grad_norm": 2.0759809017181396,
+      "learning_rate": 2.917901024719538e-05,
+      "loss": 0.8121,
+      "step": 3300
+    },
+    {
+      "epoch": 3.92,
+      "grad_norm": 2.325293779373169,
+      "learning_rate": 2.9106162886698072e-05,
+      "loss": 0.8076,
+      "step": 3400
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 0.5064914226531982,
+      "eval_runtime": 117.608,
+      "eval_samples_per_second": 26.656,
+      "eval_steps_per_second": 3.333,
+      "eval_wer": 0.23836535067816267,
+      "step": 3471
+    },
+    {
+      "epoch": 4.03,
+      "grad_norm": 1.8136922121047974,
+      "learning_rate": 2.9033315526200766e-05,
+      "loss": 0.7724,
+      "step": 3500
+    },
+    {
+      "epoch": 4.15,
+      "grad_norm": 7.821990013122559,
+      "learning_rate": 2.8961196639308436e-05,
+      "loss": 0.7879,
+      "step": 3600
+    },
+    {
+      "epoch": 4.26,
+      "grad_norm": 0.916205108165741,
+      "learning_rate": 2.888834927881113e-05,
+      "loss": 0.7907,
+      "step": 3700
+    },
+    {
+      "epoch": 4.38,
+      "grad_norm": 1.157679557800293,
+      "learning_rate": 2.8815501918313824e-05,
+      "loss": 0.7859,
+      "step": 3800
+    },
+    {
+      "epoch": 4.49,
+      "grad_norm": 1.1060450077056885,
+      "learning_rate": 2.874265455781652e-05,
+      "loss": 0.7934,
+      "step": 3900
+    },
+    {
+      "epoch": 4.61,
+      "grad_norm": 1.0524017810821533,
+      "learning_rate": 2.8669807197319215e-05,
+      "loss": 0.7887,
+      "step": 4000
+    },
+    {
+      "epoch": 4.72,
+      "grad_norm": 0.9493302702903748,
+      "learning_rate": 2.8596959836821913e-05,
+      "loss": 0.7962,
+      "step": 4100
+    },
+    {
+      "epoch": 4.84,
+      "grad_norm": 1.216361403465271,
+      "learning_rate": 2.8524840949929582e-05,
+      "loss": 0.7902,
+      "step": 4200
+    },
+    {
+      "epoch": 4.95,
+      "grad_norm": 0.858220100402832,
+      "learning_rate": 2.8451993589432277e-05,
+      "loss": 0.7875,
+      "step": 4300
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 0.4950461685657501,
+      "eval_runtime": 107.5304,
+      "eval_samples_per_second": 29.155,
+      "eval_steps_per_second": 3.645,
+      "eval_wer": 0.23864266626329855,
+      "step": 4339
+    },
+    {
+      "epoch": 5.07,
+      "grad_norm": 0.6915091872215271,
+      "learning_rate": 2.8379146228934974e-05,
+      "loss": 0.8172,
+      "step": 4400
+    },
+    {
+      "epoch": 5.19,
+      "grad_norm": 0.8395456671714783,
+      "learning_rate": 2.8306298868437668e-05,
+      "loss": 0.7719,
+      "step": 4500
+    },
+    {
+      "epoch": 5.3,
+      "grad_norm": 0.5675917863845825,
+      "learning_rate": 2.8233451507940365e-05,
+      "loss": 0.7725,
+      "step": 4600
+    },
+    {
+      "epoch": 5.42,
+      "grad_norm": 0.7588133215904236,
+      "learning_rate": 2.816060414744306e-05,
+      "loss": 0.7735,
+      "step": 4700
+    },
+    {
+      "epoch": 5.53,
+      "grad_norm": 0.7658157348632812,
+      "learning_rate": 2.8087756786945757e-05,
+      "loss": 0.7713,
+      "step": 4800
+    },
+    {
+      "epoch": 5.65,
+      "grad_norm": 0.6394296288490295,
+      "learning_rate": 2.801490942644845e-05,
+      "loss": 0.7683,
+      "step": 4900
+    },
+    {
+      "epoch": 5.76,
+      "grad_norm": 1.0770230293273926,
+      "learning_rate": 2.794206206595114e-05,
+      "loss": 0.784,
+      "step": 5000
+    },
+    {
+      "epoch": 5.88,
+      "grad_norm": 0.537818193435669,
+      "learning_rate": 2.786921470545384e-05,
+      "loss": 0.7705,
+      "step": 5100
+    },
+    {
+      "epoch": 5.99,
+      "grad_norm": 1.1256853342056274,
+      "learning_rate": 2.7796367344956533e-05,
+      "loss": 0.7947,
+      "step": 5200
+    },
+    {
+      "epoch": 6.0,
+      "eval_loss": 0.47799625992774963,
+      "eval_runtime": 102.2073,
+      "eval_samples_per_second": 30.673,
+      "eval_steps_per_second": 3.835,
+      "eval_wer": 0.23347451217667523,
+      "step": 5207
+    },
+    {
+      "epoch": 6.11,
+      "grad_norm": 2.259537935256958,
+      "learning_rate": 2.772351998445923e-05,
+      "loss": 0.7386,
+      "step": 5300
+    },
+    {
+      "epoch": 6.22,
+      "grad_norm": 1.8576114177703857,
+      "learning_rate": 2.7650672623961924e-05,
+      "loss": 0.7591,
+      "step": 5400
+    },
+    {
+      "epoch": 6.34,
+      "grad_norm": 3.029550075531006,
+      "learning_rate": 2.7578553737069594e-05,
+      "loss": 0.765,
+      "step": 5500
+    },
+    {
+      "epoch": 6.45,
+      "grad_norm": 1.3059985637664795,
+      "learning_rate": 2.750570637657229e-05,
+      "loss": 0.7484,
+      "step": 5600
+    },
+    {
+      "epoch": 6.57,
+      "grad_norm": 2.597036600112915,
+      "learning_rate": 2.7432859016074985e-05,
+      "loss": 0.7696,
+      "step": 5700
+    },
+    {
+      "epoch": 6.68,
+      "grad_norm": 1.8154231309890747,
+      "learning_rate": 2.736074012918265e-05,
+      "loss": 0.7642,
+      "step": 5800
+    },
+    {
+      "epoch": 6.8,
+      "grad_norm": 1.9362813234329224,
+      "learning_rate": 2.728789276868535e-05,
+      "loss": 0.7703,
+      "step": 5900
+    },
+    {
+      "epoch": 6.91,
+      "grad_norm": 2.5824599266052246,
+      "learning_rate": 2.7215045408188043e-05,
+      "loss": 0.7621,
+      "step": 6000
+    },
+    {
+      "epoch": 7.0,
+      "eval_loss": 0.4984392821788788,
+      "eval_runtime": 96.472,
+      "eval_samples_per_second": 32.496,
+      "eval_steps_per_second": 4.063,
+      "eval_wer": 0.23700398326022287,
+      "step": 6075
+    },
+    {
+      "epoch": 7.03,
+      "grad_norm": 1.1438063383102417,
+      "learning_rate": 2.714219804769074e-05,
+      "loss": 0.7505,
+      "step": 6100
+    },
+    {
+      "epoch": 7.14,
+      "grad_norm": 0.8449379205703735,
+      "learning_rate": 2.7069350687193434e-05,
+      "loss": 0.7626,
+      "step": 6200
+    },
+    {
+      "epoch": 7.26,
+      "grad_norm": 1.075359582901001,
+      "learning_rate": 2.699650332669613e-05,
+      "loss": 0.7503,
+      "step": 6300
+    },
+    {
+      "epoch": 7.37,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 14440011566940.16,
+      "step": 6400
+    },
+    {
+      "epoch": 7.49,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 6500
+    },
+    {
+      "epoch": 7.6,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 6600
+    },
+    {
+      "epoch": 7.72,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 6700
+    },
+    {
+      "epoch": 7.84,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 6800
+    },
+    {
+      "epoch": 7.95,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 6900
+    },
+    {
+      "epoch": 8.0,
+      "eval_loss": NaN,
+      "eval_runtime": 97.0769,
+      "eval_samples_per_second": 32.294,
+      "eval_steps_per_second": 4.038,
+      "eval_wer": 1.0,
+      "step": 6943
+    },
+    {
+      "epoch": 8.07,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7000
+    },
+    {
+      "epoch": 8.18,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7100
+    },
+    {
+      "epoch": 8.3,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7200
+    },
+    {
+      "epoch": 8.41,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7300
+    },
+    {
+      "epoch": 8.53,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7400
+    },
+    {
+      "epoch": 8.64,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7500
+    },
+    {
+      "epoch": 8.76,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7600
+    },
+    {
+      "epoch": 8.87,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7700
+    },
+    {
+      "epoch": 8.99,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7800
+    },
+    {
+      "epoch": 9.0,
+      "eval_loss": NaN,
+      "eval_runtime": 108.3909,
+      "eval_samples_per_second": 28.923,
+      "eval_steps_per_second": 3.617,
+      "eval_wer": 1.0,
+      "step": 7810
+    },
+    {
+      "epoch": 9.1,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 7900
+    },
+    {
+      "epoch": 9.22,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8000
+    },
+    {
+      "epoch": 9.33,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8100
+    },
+    {
+      "epoch": 9.45,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8200
+    },
+    {
+      "epoch": 9.56,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8300
+    },
+    {
+      "epoch": 9.68,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8400
+    },
+    {
+      "epoch": 9.79,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8500
+    },
+    {
+      "epoch": 9.91,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8600
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": NaN,
+      "eval_runtime": 107.8221,
+      "eval_samples_per_second": 29.076,
+      "eval_steps_per_second": 3.636,
+      "eval_wer": 1.0,
+      "step": 8678
+    },
+    {
+      "epoch": 10.02,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8700
+    },
+    {
+      "epoch": 10.14,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8800
+    },
+    {
+      "epoch": 10.25,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 8900
+    },
+    {
+      "epoch": 10.37,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9000
+    },
+    {
+      "epoch": 10.49,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9100
+    },
+    {
+      "epoch": 10.6,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9200
+    },
+    {
+      "epoch": 10.72,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9300
+    },
+    {
+      "epoch": 10.83,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9400
+    },
+    {
+      "epoch": 10.95,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9500
+    },
+    {
+      "epoch": 11.0,
+      "eval_loss": NaN,
+      "eval_runtime": 116.4634,
+      "eval_samples_per_second": 26.918,
+      "eval_steps_per_second": 3.366,
+      "eval_wer": 1.0,
+      "step": 9546
+    },
+    {
+      "epoch": 11.06,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9600
+    },
+    {
+      "epoch": 11.18,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9700
+    },
+    {
+      "epoch": 11.29,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9800
+    },
+    {
+      "epoch": 11.41,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 9900
+    },
+    {
+      "epoch": 11.52,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10000
+    },
+    {
+      "epoch": 11.64,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10100
+    },
+    {
+      "epoch": 11.75,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10200
+    },
+    {
+      "epoch": 11.87,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10300
+    },
+    {
+      "epoch": 11.98,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10400
+    },
+    {
+      "epoch": 12.0,
+      "eval_loss": NaN,
+      "eval_runtime": 114.432,
+      "eval_samples_per_second": 27.396,
+      "eval_steps_per_second": 3.426,
+      "eval_wer": 1.0,
+      "step": 10414
+    },
+    {
+      "epoch": 12.1,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10500
+    },
+    {
+      "epoch": 12.21,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10600
+    },
+    {
+      "epoch": 12.33,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10700
+    },
+    {
+      "epoch": 12.44,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10800
+    },
+    {
+      "epoch": 12.56,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 10900
+    },
+    {
+      "epoch": 12.67,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11000
+    },
+    {
+      "epoch": 12.79,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11100
+    },
+    {
+      "epoch": 12.91,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11200
+    },
+    {
+      "epoch": 13.0,
+      "eval_loss": NaN,
+      "eval_runtime": 107.4788,
+      "eval_samples_per_second": 29.169,
+      "eval_steps_per_second": 3.647,
+      "eval_wer": 1.0,
+      "step": 11282
+    },
+    {
+      "epoch": 13.02,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11300
+    },
+    {
+      "epoch": 13.14,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11400
+    },
+    {
+      "epoch": 13.25,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11500
+    },
+    {
+      "epoch": 13.37,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11600
+    },
+    {
+      "epoch": 13.48,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11700
+    },
+    {
+      "epoch": 13.6,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11800
+    },
+    {
+      "epoch": 13.71,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 11900
+    },
+    {
+      "epoch": 13.83,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12000
+    },
+    {
+      "epoch": 13.94,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12100
+    },
+    {
+      "epoch": 14.0,
+      "eval_loss": NaN,
+      "eval_runtime": 106.6864,
+      "eval_samples_per_second": 29.385,
+      "eval_steps_per_second": 3.674,
+      "eval_wer": 1.0,
+      "step": 12150
+    },
+    {
+      "epoch": 14.06,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12200
+    },
+    {
+      "epoch": 14.17,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12300
+    },
+    {
+      "epoch": 14.29,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12400
+    },
+    {
+      "epoch": 14.4,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12500
+    },
+    {
+      "epoch": 14.52,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12600
+    },
+    {
+      "epoch": 14.63,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12700
+    },
+    {
+      "epoch": 14.75,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12800
+    },
+    {
+      "epoch": 14.86,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 12900
+    },
+    {
+      "epoch": 14.98,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13000
+    },
+    {
+      "epoch": 15.0,
+      "eval_loss": NaN,
+      "eval_runtime": 109.7539,
+      "eval_samples_per_second": 28.564,
+      "eval_steps_per_second": 3.572,
+      "eval_wer": 1.0,
+      "step": 13018
+    },
+    {
+      "epoch": 15.09,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13100
+    },
+    {
+      "epoch": 15.21,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13200
+    },
+    {
+      "epoch": 15.32,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13300
+    },
+    {
+      "epoch": 15.44,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13400
+    },
+    {
+      "epoch": 15.56,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13500
+    },
+    {
+      "epoch": 15.67,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13600
+    },
+    {
+      "epoch": 15.79,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13700
+    },
+    {
+      "epoch": 15.9,
+      "grad_norm": NaN,
+      "learning_rate": 2.697756301296683e-05,
+      "loss": 0.0,
+      "step": 13800
+    },
+    {
+      "epoch": 16.0,
+      "eval_loss": NaN,
+      "eval_runtime": 108.5954,
+      "eval_samples_per_second": 28.869,
+      "eval_steps_per_second": 3.61,
+      "eval_wer": 1.0,
+      "step": 13886
+    },
+    {
+      "epoch": 16.0,
+      "step": 13886,
+      "total_flos": 6.139205851819624e+20,
+      "train_loss": 103989713142.67763,
+      "train_runtime": 155022.8715,
+      "train_samples_per_second": 53.744,
+      "train_steps_per_second": 0.28
+    }
+  ],
+  "logging_steps": 100,
+  "max_steps": 43350,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 50,
+  "save_steps": 500,
+  "total_flos": 6.139205851819624e+20,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f07d08380f5593b7f8241aafe43b713d22c2c6ee62cc83cc34439f35bf4f3dc
+size 4984

vocab.json

@@ -0,0 +1 @@
+{"<pad>": 0, "|": 1, "<unk>": 2, "a": 3, "b": 4, "c": 5, "d": 6, "e": 7, "f": 8, "g": 9, "h": 10, "i": 11, "j": 12, "k": 13, "l": 14, "m": 15, "n": 16, "o": 17, "p": 18, "q": 19, "r": 20, "s": 21, "t": 22, "u": 23, "v": 24, "w": 25, "x": 26, "y": 27, "z": 28, "ç": 29, "ã": 30, "à": 31, "á": 32, "â": 33, "ê": 34, "é": 35, "í": 36, "ó": 37, "ô": 38, "õ": 39, "ú": 40, "û": 41, "-": 42, "<s>": 43, "</s>": 44}