Saving weights and logs of step 1000

events.out.tfevents.1635874712.t1v-n-f6f5b6cc-w-0.906011.0.v2 → events.out.tfevents.1635877753.t1v-n-f6f5b6cc-w-0.912674.0.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5107e915386751f93dcbcf73081df41a8318804cb05953320b8e4251edac6e9a
-size 2209127
+oid sha256:95f27c4c3af3da4607838029ca94c7ecc087d9a100294e7a744433c72e8e97f1
+size 147207

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:00532437064025b9e8e11ab637b7e5821e52d0d92bdac36ad7b9c4ec52c54aee
+oid sha256:9b32778e6c2b848bdc5d724eb1774483f64902c76609aa8ca75cfc9157dc711b
 size 497764120

run.sh

@@ -11,11 +11,11 @@ python run_clm_flax.py \
     --block_size="512" \
     --per_device_train_batch_size="8" \
     --per_device_eval_batch_size="8" \
-    --learning_rate="1e-4" \
-    --warmup_steps="100" \
+    --learning_rate="3e-5" \
+    --warmup_steps="500" \
     --adam_beta1="0.9" --adam_beta2="0.98" --weight_decay="0.01" \
     --overwrite_output_dir \
-    --num_train_epochs="30" \
+    --num_train_epochs="10" \
     --logging_steps="500" \
     --save_steps="1000" \
     --eval_steps="1000" \