Training in progress, step 100

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "juancopi81/whisper-medium-es-train-valid-bs-64",
+  "_name_or_path": "juancopi81/whisper-medium-es",
   "activation_dropout": 0.0,
   "activation_function": "gelu",
   "architectures": [

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:619bd4c8a11c4fe601524a7c0beb23fb82404bd4aa82d1b0783a1aa0557034b3
+oid sha256:35e800d4522326115904b878c2c708d6b6ab3d9332c3bcf82017c8220a7db8ba
 size 1527847357

run.sh

@@ -1,24 +1,24 @@
 deepspeed run_speech_recognition_seq2seq_streaming.py \
 --deepspeed="ds_config.json" \
---model_name_or_path="juancopi81/whisper-medium-es-train-valid-bs-64" \
+--model_name_or_path="juancopi81/whisper-medium-es" \
 --dataset_name="mozilla-foundation/common_voice_11_0" \
 --dataset_config_name="es" \
 --language="spanish" \
 --train_split_name="train+validation" \
 --eval_split_name="test" \
 --model_index_name="Whisper Small Spanish" \
---max_steps="500" \
+--max_steps="100" \
 --output_dir="./" \
 --per_device_train_batch_size="64" \
 --per_device_eval_batch_size="32" \
 --logging_steps="25" \
---learning_rate="5e-7" \
---warmup_steps="100" \
+--learning_rate="3e-5" \
+--warmup_steps="10" \
 --evaluation_strategy="steps" \
---eval_steps="500" \
+--eval_steps="100" \
 --save_strategy="steps" \
---save_steps="500" \
---generation_max_length="225" \
+--save_steps="100" \
+--generation_max_length="5" \
 --length_column_name="input_length" \
 --max_duration_in_seconds="30" \
 --text_column_name="sentence" \

runs/Dec16_21-53-30_132-145-140-45/1671227636.32413/events.out.tfevents.1671227636.132-145-140-45.989061.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a219b172a2b71d6fbc6cfbafa84f4b288a28e4e4115c5cf67b3f756bd21ea25
+size 5880

runs/Dec16_21-53-30_132-145-140-45/events.out.tfevents.1671227636.132-145-140-45.989061.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d453887acd46dafa47a27f85d23408fe31afa5092146fef1edfd37f4d28480bf
+size 5209

tokenizer_config.json

@@ -19,7 +19,7 @@
   },
   "errors": "replace",
   "model_max_length": 1024,
-  "name_or_path": "juancopi81/whisper-medium-es-train-valid-bs-64",
+  "name_or_path": "juancopi81/whisper-medium-es",
   "pad_token": null,
   "processor_class": "WhisperProcessor",
   "return_attention_mask": false,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:24d4aa81ef8f537670affe4d5b2c5869ac70b558e4107f25cdf01c52465bbe29
+oid sha256:7a0cc955785ae0c566469a79ef4a6f8cab101e3beac5ec6c5745bd6ef17fc6a7
 size 4667