Training in progress, step 500

config.json

@@ -3,7 +3,6 @@
   "architectures": [
     "SpeechEncoderDecoderModel"
   ],
-  "bos_token_id": 50256,
   "decoder": {
     "_name_or_path": "gpt2-medium",
     "activation_function": "gelu_new",

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a77c253742ba1d1d65de2970c1f75ab1bcfc074438a088bdeef850d1f8880967
+oid sha256:c5b9ee785de0e4c8d296e99160eba9703ba7b5a30f6d9c758974ff5709e7a7af
 size 3210536233

run_librispeech.sh

@@ -16,7 +16,7 @@ CUDA_VISIBLE_DEVICES=0 python run_speech_recognition_seq2seq.py \
 	--gradient_accumulation_steps="2" \
 	--generation_max_length="40" \
 	--generation_num_beams="1" \
-	--learning_rate="3e-4" \
+	--learning_rate="1e-4" \
 	--warmup_steps="500" \
 	--evaluation_strategy="steps" \
 	--text_column_name="text" \

runs/Feb11_14-18-53_sanchit--v100/1644589174.7624567/events.out.tfevents.1644589174.sanchit--v100.346942.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a234beaa2ec98b2905c8a2b66043e0158a41200ea72b5d2e72003f85271d8c7
+size 4964

runs/Feb11_14-18-53_sanchit--v100/events.out.tfevents.1644589174.sanchit--v100.346942.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1674ac90d465b462ff4dab3427086efbb2e68a14e96459767a909c2632cc943c
+size 87604

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5581e74fa317eb5734b81037a72a51c888100315cb839743cf07ae7d5596e61a
+oid sha256:8640bd32aac20e5c90224ed112830c3024043dfa9e15a5d950e20f542d170e64
 size 3119