Training in progress, step 200

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "openai/whisper-large-v2",
+  "_name_or_path": "p4b/whisper-large-v2-lv",
   "activation_dropout": 0.0,
   "activation_function": "gelu",
   "architectures": [

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7a115ff886475c9aeba1f4b5db24a576404bc496725b85ec2b92809ea3124e7a
-size 3086885937
+oid sha256:6221c007deae46a3c7638e49b86491a5170532af9fba7509175501769a069de0
+size 3086785087

run2.sh

@@ -0,0 +1,43 @@
+deepspeed run_speech_recognition_seq2seq_streaming.py \
+    --deepspeed="ds_config.json" \
+    --model_name_or_path="p4b/whisper-large-v2-lv" \
+    --dataset_name="mozilla-foundation/common_voice_11_0" \
+    --dataset_config_name="lv" \
+    --language="latvian" \
+    --train_split_name="train+validation" \
+    --eval_split_name="test" \
+    --model_index_name="Whisper Large-v2 Latvian" \
+    --max_steps="900" \
+    --output_dir="./" \
+    --per_device_train_batch_size="32" \
+    --gradient_accumulation_steps="2" \
+    --per_device_eval_batch_size="32" \
+    --logging_steps="25" \
+    --learning_rate="1e-7" \
+    --lr_scheduler_type="cosine" \
+    --warmup_steps="100" \
+    --evaluation_strategy="steps" \
+    --eval_steps="200" \
+    --save_strategy="steps" \
+    --save_steps="200" \
+    --generation_max_length="225" \
+    --length_column_name="input_length" \
+    --max_duration_in_seconds="30" \
+    --text_column_name="sentence" \
+    --freeze_feature_encoder="False" \
+    --freeze_encoder="False" \
+    --report_to="tensorboard" \
+    --metric_for_best_model="wer" \
+    --greater_is_better="False" \
+    --load_best_model_at_end \
+    --gradient_checkpointing \
+    --fp16 \
+    --overwrite_output_dir \
+    --do_train \
+    --do_eval \
+    --predict_with_generate \
+    --do_normalize_eval \
+    --use_auth_token \
+    --push_to_hub \
+    --torch_compile="True" \
+    --local_rank=0

runs/Dec19_09-23-32_nipa2022-65403/1671409649.396783/events.out.tfevents.1671409649.nipa2022-65403.20673.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb0f26a93ceec45a313b689d30a026934641ad8461251f435399821468811ee2
+size 5884

runs/Dec19_09-23-32_nipa2022-65403/events.out.tfevents.1671409649.nipa2022-65403.20673.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8dcff521bf933ebe64d9cda1274dba46b3aff08f7a273c6b42eb014ae38dc8c
+size 4286

runs/Dec19_09-48-52_nipa2022-65403/1671411014.4457693/events.out.tfevents.1671411014.nipa2022-65403.3427.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcf44527ec2f968d95208ac8f04ab2a008abff1dbc03872258edd74f0b8bc603
+size 5884

runs/Dec19_09-48-52_nipa2022-65403/events.out.tfevents.1671411014.nipa2022-65403.3427.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fd3f4b4cee94f2651c27cee567818ed46a42cdedb1d770deb6c411f7ae53abf
+size 5845

tokenizer_config.json

@@ -19,7 +19,7 @@
   },
   "errors": "replace",
   "model_max_length": 1024,
-  "name_or_path": "openai/whisper-large-v2",
+  "name_or_path": "p4b/whisper-large-v2-lv",
   "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 sha256:3799dc22368507dfe5fa8baa5ff4c56435981c120180031622d237adfa2f6550
+oid sha256:5515b192c2ca73529b118c5b04d885e8951cd2f2e0b013ca9552fad66d1fe844
 size 4411