Saving weights and logs of step 20000

events.out.tfevents.1674498755.t1v-n-9758a16f-w-0.1006414.0.v2 → events.out.tfevents.1674544407.t1v-n-9758a16f-w-0.1066395.0.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:01991d4a0dd7dca386b7c2994710e7e56d51d6a650876867ced7ea350f6631ef
-size 11899197
+oid sha256:f38f6d9fa18938fd0ba282b44d7440a8507ecea5217448ec00125ef5f8058f55
+size 2950146

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4219e0d6389719e5379375739eea2edbb78f4401322e909a0ef14c4a9ef52dfe
+oid sha256:d5e88922a1080d678c723871d9177d422ba926f7810feb7781c4395336ee29a4
 size 1625384886

start_train.sh

@@ -7,7 +7,7 @@ python run_bart_dlm_flax.py \
     --max_seq_length="1024" \
     --per_device_train_batch_size="4" \
     --per_device_eval_batch_size="4" \
-    --learning_rate="4e-4" \
+    --learning_rate="4e-5" \
     --weight_decay="0.01" \
     --warmup_steps="10000" \
     --overwrite_output_dir \