Training in progress, step 200

.gitattributes

@@ -32,3 +32,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+cache/common_voice_walkie/ru/6.1.0/da0b466819242bff763a2319e08f49c101cf8efe6dae0d7468c3536a40b1addd/tmpf0pfrlg9 filter=lfs diff=lfs merge=lfs -text

cache/common_voice_walkie/ru/6.1.0/da0b466819242bff763a2319e08f49c101cf8efe6dae0d7468c3536a40b1addd/cache-4ec5d413f38fa76b.arrow

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be2c8116543fdb45df72a764c42670f1b2ea5f5440c4e59c9aaa7bb1afb4b778
+size 2773103360

cache/common_voice_walkie/ru/6.1.0/da0b466819242bff763a2319e08f49c101cf8efe6dae0d7468c3536a40b1addd/cache-580592b2fb57127e.arrow

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:962039a3de3e4deb6206cd9b228180d23138d1f8d824ca89ba81e1db89cd8d94
+size 280592

cache/common_voice_walkie/ru/6.1.0/da0b466819242bff763a2319e08f49c101cf8efe6dae0d7468c3536a40b1addd/cache-aa87c5bc3604ea74.arrow

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50aba9e485e22ed0f82d6e42a16d474ce3f2877016b8579a3fabe2c399f00157
+size 33048025696

cache/common_voice_walkie/ru/6.1.0/da0b466819242bff763a2319e08f49c101cf8efe6dae0d7468c3536a40b1addd/tmpf0pfrlg9

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:debee7770f9aff51ae42f4edf49a4b9e383f9f4d9c6560a01b92e0eb9bcde35c
+size 3842117464

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06d41119ed8ad3928ee7ca7219544b2c601fdde4ae17701d9e382de4615843db
+oid sha256:de9477105cad7c3c04c8d72d24b68c5edc9f6011efa2ffd0c035f5cb0db38aa6
 size 6173655480

run.sh

@@ -15,9 +15,9 @@ python run_speech_recognition_seq2seq_streaming.py \
 	--learning_rate="1e-5" \
 	--warmup_steps="500" \
 	--evaluation_strategy="steps" \
-	--eval_steps="1000" \
+	--eval_steps="200" \
 	--save_strategy="steps" \
-	--save_steps="1000" \
+	--save_steps="200" \
 	--generation_max_length="225" \
 	--length_column_name="input_length" \
 	--max_duration_in_seconds="30" \
@@ -36,6 +36,8 @@ python run_speech_recognition_seq2seq_streaming.py \
 	--do_eval \
 	--predict_with_generate \
 	--do_normalize_eval \
+	--do_lower_case \
+	--do_remove_punctuation \
 	--streaming="False" \
 	--use_auth_token \
 	--push_to_hub

run_speech_recognition_seq2seq_streaming.py

@@ -23,6 +23,7 @@ with 🤗 Datasets' streaming mode.
 import logging
 import os
 import sys
+import re
 from dataclasses import dataclass, field
 from typing import Any, Dict, List, Optional, Union
 
@@ -477,6 +478,16 @@ def main():
             else raw_datasets["eval"].select(range(data_args.max_eval_samples))
         )
 
+    def my_normalize(text):
+        valid_tokens = "ж - п и щ я о ш ц ъ д а й х э г т з ч ю б н ф ы е ё к р в ь с л у м"
+        valid_tokens = set([x.lower() for x in valid_tokens])
+        # The caret in the character class ([^) means match anything but
+        invalid_chars_regex = f"[^\s{re.escape(''.join(set(valid_tokens)))}]"
+        text = text.lower() if do_lower_case else text.upper()
+        text = re.sub(invalid_chars_regex, " ", text)
+        text = re.sub("\s+", " ", text).strip()
+        return text
+
     def prepare_dataset(batch):
         # process audio
         sample = batch[audio_column_name]
@@ -489,6 +500,7 @@ def main():
         input_str = batch[text_column_name].lower() if do_lower_case else batch[text_column_name]
         if do_remove_punctuation:
             input_str = normalizer(input_str).strip()
+            input_str = my_normalize(input_str).strip()
         batch["labels"] = tokenizer(input_str).input_ids
         return batch
 

runs/Mar07_17-04-19_23e47d4ed5a0/events.out.tfevents.1678208691.23e47d4ed5a0.1578177.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2339fa7b3a4df20c5631216ae0e9097d62b73b9b954b968e0cd964b829bc53bb
-size 17805
+oid sha256:8e0f5e3772e57e55a0ed5979498fd8fc21c659ef8c330a7125ead7721caf6bf7
+size 22672

runs/Mar12_08-09-39_23e47d4ed5a0/1678613181.4145474/events.out.tfevents.1678613181.23e47d4ed5a0.410.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5773c502d21c61df4d71265495f122e8e78cc1356f28c143b2bdfc150b945744
+size 6027

runs/Mar12_08-09-39_23e47d4ed5a0/events.out.tfevents.1678613181.23e47d4ed5a0.410.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47204d466cacdb28ccdb97755f2aff49e3f3bc93a97167d9dd55097a88d76070
+size 6181

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7c72dfc4edc78d738167d1d68154d3eb5f654ca466d6c04b4318f13d108c8f6f
+oid sha256:b7d55acb6843dac8089f2fa21146da1a4542b214f1e31b9f55026e7d8f400336
 size 3707