Training in progress, step 20

checkpoint-20/config.json

@@ -0,0 +1,115 @@
+{
+  "_name_or_path": "KBLab/wav2vec2-large-voxrex",
+  "activation_dropout": 0.1,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.0,
+  "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": 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": 0.0,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.0,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.0,
+  "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": 64,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.25,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_min_space": 1,
+  "mask_time_other": 0.0,
+  "mask_time_prob": 0.75,
+  "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": 31,
+  "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": "4.17.0.dev0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 34,
+  "xvector_output_dim": 512
+}

checkpoint-20/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:075df4bf25f6d34069d73115a8c7331da68aa4a7dd3f23353e295886bc03506a
+size 2490337361

checkpoint-20/preprocessor_config.json

@@ -0,0 +1,9 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

checkpoint-20/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9b6100773ca3a4687e69ef744a414602f964603f13c71a09f8727f73dd051f1
+size 1262063089

checkpoint-20/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73e968b5fb75f3d028c5b87fc29d361b46586d839eb1fc406be3e77c69542778
+size 14567

checkpoint-20/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8d1040cd70f91f849688103143e8c9d631ff10c4acee0b2f4d6ff9e27727776
+size 559

checkpoint-20/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bee97b3839fd1b431d21567a2339671d799c163d5402ee322afd861513750b5
+size 623

checkpoint-20/trainer_state.json

@@ -0,0 +1,37 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.26490066225165565,
+  "global_step": 20,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.9999999999999996e-05,
+      "loss": 13.2048,
+      "step": 10
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.46938775510204e-05,
+      "loss": 26.5657,
+      "step": 20
+    },
+    {
+      "epoch": 0.26,
+      "eval_loss": 17.199691772460938,
+      "eval_runtime": 5.248,
+      "eval_samples_per_second": 18.674,
+      "eval_steps_per_second": 0.762,
+      "eval_wer": 1.0,
+      "step": 20
+    }
+  ],
+  "max_steps": 750,
+  "num_train_epochs": 10,
+  "total_flos": 4.155165485184e+17,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-20/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:470abf25a211ef5ec3a9821c884f2749839bee5e95f633b168d26f90c1809409
+size 3055

old_run.sh

@@ -6,11 +6,13 @@ python old_run_asr_ctc.py \
 	--eval_split_name="test,None" \
 	--output_dir="./" \
 	--overwrite_output_dir \
-	--num_train_epochs="3" \
+	--num_train_epochs="10" \
 	--per_device_train_batch_size="32" \
 	--per_device_eval_batch_size="32" \
 	--gradient_accumulation_steps="4" \
-	--learning_rate="7.5e-4" \
+	--learning_rate="7.5e-5" \
+    --max_train_samples="10000" \
+    --max_eval_samples="100" \
 	--warmup_ratio="0.0" \
 	--length_column_name="input_length" \
 	--evaluation_strategy="steps" \

old_run_asr_ctc.py

@@ -525,7 +525,7 @@ def main():
                     )
                     new_dataset_sampling_rate = new_dataset["eval"].features[data_args.audio_column_name].sampling_rate
                     if new_dataset_sampling_rate != dataset_sampling_rate:
-                        print(f"New dataset sampling rate casted from {dataset_sampling_rate} to {dataset_sampling_rate}")
+                        print(f"New dataset sampling rate casted from {new_dataset_sampling_rate} to {dataset_sampling_rate}")
                         new_dataset["eval"] = new_dataset["eval"].cast_column(
                             data_args.audio_column_name, datasets.features.Audio(sampling_rate=dataset_sampling_rate)
                         )

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4cfc19f8f8b247c6b5050f32c8b2bf68b9bdc2656dafb45faf51286ab02eec5e
+oid sha256:e9b6100773ca3a4687e69ef744a414602f964603f13c71a09f8727f73dd051f1
 size 1262063089

run.sh

@@ -6,20 +6,22 @@ python run_speech_recognition_ctc.py \
 	--eval_split_name="test" \
 	--output_dir="./" \
 	--overwrite_output_dir \
-	--num_train_epochs="3" \
+	--num_train_epochs="10" \
 	--per_device_train_batch_size="32" \
 	--per_device_eval_batch_size="32" \
 	--gradient_accumulation_steps="4" \
-	--learning_rate="7.5e-" \
+	--learning_rate="7.5e-5" \
+	--max_train_samples="10000" \
+    --max_eval_samples="100" \
 	--warmup_ratio="0.02" \
 	--length_column_name="input_length" \
 	--evaluation_strategy="steps" \
 	--save_strategy="steps" \
-	--eval_steps="100" \
-	--save_steps="100" \
+	--eval_steps="20" \
+	--save_steps="20" \
 	--text_column_name="sentence" \
 	--chars_to_ignore , ? . ! \- \; \: \" “ % ‘ ” � — ’ … – \
-	--logging_steps="20" \
+	--logging_steps="10" \
 	--layerdrop="0.0" \
 	--activation_dropout="0.1" \
 	--save_total_limit="2" \
@@ -39,7 +41,7 @@ python run_speech_recognition_ctc.py \
 python run_speech_recognition_ctc.py \
 	--dataset_name="mozilla-foundation/common_voice_8_0" \
 	--model_name_or_path="KBLab/wav2vec2-large-voxrex" \
-	--dataset_config_name="distant_channel" \
+	--dataset_config_name="sv-SE" \
 	--train_split_name="train+validation" \
 	--eval_split_name="test" \
 	--output_dir="./" \

run_speech_recognition_ctc.py

@@ -476,6 +476,9 @@ def main():
                 f"{', '.join(raw_datasets['train'].column_names)}."
             )
 
+        dataset_frequency = raw_datasets["train"].features[data_args.audio_column_name].sampling_rate
+        print(f"Dataset sampling rate: {dataset_frequency}")
+
         if data_args.text_column_name not in raw_datasets["train"].column_names:
             raise ValueError(
                 f"--text_column_name {data_args.text_column_name} not found in dataset '{data_args.dataset_name}'. "
@@ -528,12 +531,18 @@ def main():
         #     else:
         #         logging.warning(f"{dataset_name} {dataset_config_name} eval not loaded as split is {eval_split_name}")
 
-        raw_datasets["eval"] = load_dataset(
-            data_args.dataset_name,
-            data_args.dataset_config_name,
-            split=data_args.eval_split_name,
-            use_auth_token=data_args.use_auth_token,
-        )
+        try:
+            raw_datasets["eval"] = load_dataset(
+                data_args.dataset_name,
+                data_args.dataset_config_name,
+                split=data_args.eval_split_name,
+                use_auth_token=data_args.use_auth_token,
+            )
+        except ValueError:
+            split_dataset = raw_datasets["train"].train_test_split(test_size=0.1, seed=42)
+            raw_datasets["eval"] = split_dataset["test"]
+            print(raw_datasets["eval"])
+            print("Sampled from training set")
 
         if data_args.max_eval_samples is not None:
             raw_datasets["eval"] = raw_datasets["eval"].select(range(data_args.max_eval_samples))

special_tokens_map.json

@@ -1 +1 @@
-{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]", "additional_special_tokens": [{"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}]}
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]", "additional_special_tokens": [{"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}]}

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:962dbb61fb0395a025f0dae3475a50ad9719299753b3a01bede5cb391a4aca78
+oid sha256:470abf25a211ef5ec3a9821c884f2749839bee5e95f633b168d26f90c1809409
 size 3055