Training in progress, step 1500

ex3/vocab.json

@@ -0,0 +1 @@
+{"a": 1, "b": 2, "c": 3, "d": 4, "e": 5, "f": 6, "g": 7, "h": 8, "i": 9, "j": 10, "k": 11, "l": 12, "m": 13, "n": 14, "o": 15, "p": 16, "q": 17, "r": 18, "s": 19, "t": 20, "u": 21, "v": 22, "w": 23, "x": 24, "y": 25, "z": 26, "\u00e5": 27, "\u00e6": 28, "\u00f8": 29, "|": 0, "[UNK]": 30, "[PAD]": 31}

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1db06c6f208919999109fb54084bd07850ef535e6dbd5edad7741b1b78a55c7b
+oid sha256:f6edf3b6d111f124f397bb09783b024609d2532d55c30be41df0e5b870da8916
 size 1262063089

run.sh

@@ -8,7 +8,7 @@ python run_speech_recognition_ctc.py \
         --per_device_train_batch_size="16" \
         --per_device_eval_batch_size="16" \
         --gradient_accumulation_steps="4" \
-        --learning_rate="3e-5" \
+        --learning_rate="7.5e-5" \
         --warmup_steps="500" \
         --length_column_name="input_length" \
         --evaluation_strategy="steps" \

run_ex1.sh

@@ -0,0 +1,36 @@
+python run_speech_recognition_ctc.py \
+        --dataset_name="NbAiLab/NPSC" \
+        --model_name_or_path="facebook/wav2vec2-xls-r-300m" \
+        --dataset_config_name="16K_mp3" \
+        --output_dir="./" \
+        --overwrite_output_dir \
+        --num_train_epochs="20" \
+        --per_device_train_batch_size="16" \
+        --per_device_eval_batch_size="16" \
+        --gradient_accumulation_steps="4" \
+        --learning_rate="3e-5" \
+        --warmup_steps="500" \
+        --length_column_name="input_length" \
+        --evaluation_strategy="steps" \
+        --text_column_name="text" \
+        --save_steps="250" \
+        --eval_steps="250" \
+        --logging_steps="50" \
+        --layerdrop="0.0" \
+        --activation_dropout="0.1" \
+        --save_total_limit="3" \
+        --freeze_feature_encoder \
+        --feat_proj_dropout="0.0" \
+        --mask_time_prob="0.75" \
+        --mask_time_length="10" \
+        --mask_feature_prob="0.25" \
+        --mask_feature_length="64" \
+	--chars_to_ignore , ? . ! \- \; \: \" “ % ‘ ” � — ’ … – _ \\ + \# / \
+        --gradient_checkpointing \
+        --use_auth_token \
+        --fp16 \
+        --group_by_length \
+        --do_train --do_eval \
+        --push_to_hub \
+        --preprocessing_num_workers="8"
+

run_ex3.sh

@@ -0,0 +1,36 @@
+python run_speech_recognition_ctc.py \
+        --dataset_name="NbAiLab/NPSC" \
+        --model_name_or_path="facebook/wav2vec2-xls-r-300m" \
+        --dataset_config_name="16K_mp3" \
+        --output_dir="./ex3" \
+        --overwrite_output_dir \
+        --num_train_epochs="20" \
+        --per_device_train_batch_size="16" \
+        --per_device_eval_batch_size="16" \
+        --gradient_accumulation_steps="4" \
+        --learning_rate="7.5e-5" \
+        --warmup_steps="500" \
+        --length_column_name="input_length" \
+        --evaluation_strategy="steps" \
+        --text_column_name="text" \
+        --save_steps="250" \
+        --eval_steps="250" \
+        --logging_steps="50" \
+        --layerdrop="0.1" \
+        --activation_dropout="0.1" \
+        --save_total_limit="3" \
+        --freeze_feature_encoder \
+        --feat_proj_dropout="0.0" \
+        --mask_time_prob="0.75" \
+        --mask_time_length="10" \
+        --mask_feature_prob="0.25" \
+        --mask_feature_length="64" \
+	--chars_to_ignore , ? . ! \- \; \: \" “ % ‘ ” � — ’ … – _ \\ + \# / \
+        --gradient_checkpointing \
+        --use_auth_token \
+        --fp16 \
+        --group_by_length \
+        --do_train --do_eval \
+        --push_to_hub \
+        --preprocessing_num_workers="8"
+

runs/Jan31_11-10-54_ficino/events.out.tfevents.1643623929.ficino.269307.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:60f37df178c0c2b78acfd6de36eefa08f745e7ff73bedadfd56a1b79d1977423
-size 10155
+oid sha256:72a77bf0403a71bedf47a4a64b53031e1d7f37627570e4ea9710a52d44eed7ca
+size 11258