Saving weights and logs of epoch 1

events.out.tfevents.1626345259.t1v-n-b95d739e-w-0.658394.3.v2 → events.out.tfevents.1626359445.t1v-n-b95d739e-w-0.675417.3.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1294d543e4ae77c5248e8ccfeb10573138db2a3277ebbd574ede422f3c7186b7
-size 1922237
+oid sha256:d1fa8e21e1e52124489213ba47f4806548e984b1ffc47abba1ceb0accf4a5cd0
+size 78623

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef08c1f27da650b5f25e6b055b8df64ba4101f1d30961f70bb8f87399e293976
+oid sha256:5329996fb4d4a17b9193268981cb1c0303ba84ec4d2ed3b5e077c324b5c9db1f
 size 497764120

nohup.out

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:367e84b702ad2b2794d2422b86ceca88a2f91a16dadde7061bbce8e7ab2f7bc6
-size 1207226
+oid sha256:ea7f61405e3a3981ba620f3a082fa5d6dab3ea3a9883e2df6660e31f8b7d8e5f
+size 117808

run.sh

@@ -9,10 +9,10 @@ python3 run_clm_flax.py \
     --do_eval \
     --block_size="512" \
     --preprocessing_num_workers="64" \
-    --weight_decay="0.0" \
+    --weight_decay="0.01" \
     --per_device_train_batch_size="64" \
     --per_device_eval_batch_size="64" \
-    --learning_rate="2e-4" \
+    --learning_rate="1e-4" \
     --warmup_steps="1000" \
     --overwrite_output_dir \
     --num_train_epochs="50" \