Model save

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "hf-test/xls-r-dummy",
+  "_name_or_path": "facebook/wav2vec2-xls-r-300m",
   "activation_dropout": 0.0,
   "adapter_kernel_size": 3,
   "adapter_stride": 2,
@@ -11,23 +11,35 @@
   "attention_dropout": 0.0,
   "bos_token_id": 1,
   "classifier_proj_size": 256,
-  "codevector_dim": 256,
+  "codevector_dim": 768,
   "contrastive_logits_temperature": 0.1,
-  "conv_bias": false,
+  "conv_bias": true,
   "conv_dim": [
-    32,
-    32,
-    32
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
   ],
   "conv_kernel": [
-    8,
-    8,
-    8
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
   ],
   "conv_stride": [
-    4,
-    4,
-    4
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
   ],
   "ctc_loss_reduction": "mean",
   "ctc_zero_infinity": false,
@@ -42,31 +54,30 @@
   "final_dropout": 0.0,
   "hidden_act": "gelu",
   "hidden_dropout": 0.0,
-  "hidden_dropout_prob": 0.1,
-  "hidden_size": 16,
+  "hidden_size": 1024,
   "initializer_range": 0.02,
-  "intermediate_size": 20,
+  "intermediate_size": 4096,
   "layer_norm_eps": 1e-05,
   "layerdrop": 0.0,
-  "mask_feature_length": 10,
+  "mask_feature_length": 64,
   "mask_feature_min_masks": 0,
-  "mask_feature_prob": 0.0,
+  "mask_feature_prob": 0.1,
   "mask_time_length": 10,
   "mask_time_min_masks": 2,
-  "mask_time_prob": 0.05,
+  "mask_time_prob": 0.3,
   "model_type": "wav2vec2",
   "num_adapter_layers": 3,
-  "num_attention_heads": 2,
+  "num_attention_heads": 16,
   "num_codevector_groups": 2,
   "num_codevectors_per_group": 320,
-  "num_conv_pos_embedding_groups": 2,
-  "num_conv_pos_embeddings": 16,
-  "num_feat_extract_layers": 3,
-  "num_hidden_layers": 4,
-  "num_negatives": 10,
-  "output_hidden_size": 16,
+  "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": 38,
-  "proj_codevector_dim": 256,
+  "proj_codevector_dim": 768,
   "tdnn_dilation": [
     1,
     2,

preprocessor_config.json

@@ -3,7 +3,7 @@
   "feature_extractor_type": "Wav2Vec2FeatureExtractor",
   "feature_size": 1,
   "padding_side": "right",
-  "padding_value": 0.0,
-  "return_attention_mask": false,
+  "padding_value": 0,
+  "return_attention_mask": true,
   "sampling_rate": 16000
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f1900fda595fc11d224093d778614c0532b93a9ba52d38f0313c2cb347dfff4
-size 143078
+oid sha256:bc548f439611bd98eab1f2cbb578ef3fd4344acf3ed24a78ee45778dbdd9db60
+size 1262091761

run-test.shs

@@ -0,0 +1,25 @@
+python run_speech_recognition_ctc.py \
+	--dataset_name="common_voice" \
+	--model_name_or_path="hf-test/xls-r-dummy" \
+	--dataset_config_name="tr" \
+	--output_dir="./" \
+	--overwrite_output_dir \
+	--num_train_epochs="5" \
+	--per_device_train_batch_size="16" \
+	--gradient_accumulation_steps="2" \
+	--learning_rate="3e-4" \
+	--warmup_steps="500" \
+	--evaluation_strategy="steps" \
+	--text_column_name="sentence" \
+	--length_column_name="input_length" \
+	--save_steps="400" \
+	--eval_steps="100" \
+	--layerdrop="0.0" \
+	--save_total_limit="3" \
+	--freeze_feature_encoder \
+	--gradient_checkpointing \
+	--chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
+	--fp16 \
+	--group_by_length \
+	--push_to_hub \
+	--do_train --do_eval 

run.sh

@@ -1,25 +1,29 @@
 python run_speech_recognition_ctc.py \
 	--dataset_name="common_voice" \
-	--model_name_or_path="hf-test/xls-r-dummy" \
+	--model_name_or_path="facebook/wav2vec2-xls-r-300m" \
 	--dataset_config_name="tr" \
 	--output_dir="./" \
 	--overwrite_output_dir \
-	--num_train_epochs="5" \
-	--per_device_train_batch_size="16" \
-	--gradient_accumulation_steps="2" \
-	--learning_rate="3e-4" \
+	--num_train_epochs="2" \
+	--per_device_train_batch_size="32" \
+	--gradient_accumulation_steps="1" \
+	--learning_rate="5e-4" \
 	--warmup_steps="500" \
 	--evaluation_strategy="steps" \
 	--text_column_name="sentence" \
-	--length_column_name="input_length" \
-	--save_steps="400" \
-	--eval_steps="100" \
+	--save_steps="500" \
+	--eval_steps="500" \
+	--logging_steps="1" \
 	--layerdrop="0.0" \
-	--save_total_limit="3" \
-	--freeze_feature_encoder \
-	--gradient_checkpointing \
+	--eval_metrics wer cer \
+	--save_total_limit="1" \
+	--mask_time_prob="0.3" \
+	--mask_time_length="10" \
+	--mask_feature_prob="0.1" \
+	--mask_feature_length="64" \
 	--chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
 	--fp16 \
 	--group_by_length \
 	--push_to_hub \
-	--do_train --do_eval 
+	--do_train --do_eval \
+	--gradient_checkpointing 

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}]}
+{"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}]}

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fb45223fb5a36e43839ad3901144a0cc22ece7adfab348b54d71e1678afd9fca
+oid sha256:c741e2d2d858980a8cfdf104cef10aacd51be9a34f3f0fb299372de4f16f9e54
 size 2991