Saving weights and logs of step 10000

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:266f095daf4b555de3d86b59901bbee1743abbd8adfe0f42efe2592c9cb5024c
+oid sha256:c7f89852d3a6a38756fae270a2926fabccc5ded47ed03b2bdf72323d30b48553
 size 497764120

runs/events.out.tfevents.1642099734.t1v-n-42145f73-w-0.2317757.0.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f8bf958ee051c3c421e09d54c259abc1828f8068836d91ad1d31ba30183ddfbc
-size 14882567
+oid sha256:dbbf5988858cf76da919fb1758c11b0ecc821048cb24a60adc8a53b7f332f86b
+size 15255487

runs/events.out.tfevents.1642208918.t1v-n-42145f73-w-0.2567321.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f34abfda811639f3d9abfb3c2ce67f5fda43568d82efbc314a084ecb090ebf6
+size 1471449

start_train.sh

@@ -12,7 +12,7 @@ python3 run_clm_flax.py \
     --block_size="512" \
     --per_device_train_batch_size="32" \
     --per_device_eval_batch_size="32" \
-    --preprocessing_num_workers="64" \
+    --preprocessing_num_workers="1" \
     --learning_rate="5e-3" \
     --warmup_ratio="0.01" \
     --cosine_decay \