Saving weights and logs of step 5001

events.out.tfevents.1627322413.t1v-n-358ff5d1-w-0.279972.3.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c9e15479082e44c4261173873c3a16d551975e2e10b90889be5c54a2be6fb9f9
-size 3324045
+oid sha256:6ae095ce195c7b2c0098000b3fb8dd4987fb0b185b434ac9c30a660382fdc005
+size 3622325

events.out.tfevents.1627408712.t1v-n-358ff5d1-w-0.356869.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac4f12c4cb39a097afa86e70d7c9722a7e4da9c1da90a62c3c2c93e7e2aabae6
+size 735757

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:889cd42996671090d611be12498ee5397feee4d117dc9eeb7fbcf1adf057dec3
+oid sha256:3e11efa6140818e1689b29e00c7c52aa5db87ea7638cf2ac4c245bc49951ebba
 size 1100762015

run_streaming.sh

@@ -9,17 +9,17 @@
     --per_device_train_batch_size="32" \
     --per_device_eval_batch_size="32" \
     --learning_rate="8e-3" \
-    --warmup_steps="2000" \
+    --warmup_steps="5000" \
     --overwrite_output_dir \
     --cache_dir /mnt/disks/flaxdisk/cache/ \
     --num_train_epochs="5" \
     --adam_beta1="0.9" \
     --adam_beta2="0.98" \
     --logging_steps="500" \
-    --num_train_steps="100000" \
+    --num_train_steps="500000" \
     --num_eval_samples="5000" \
-    --save_steps="2500" \
-    --eval_steps="2500" \
+    --save_steps="5000" \
+    --eval_steps="5000" \
     --preprocessing_num_workers 96 \
     --adafactor \
     --push_to_hub