Saving weights and logs of step 18000

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ed68bf4bf2ba245a90ae31d71a56c0f85d1ef7665f0748bd10826c688e5de825
+oid sha256:bfe1b36176c2200bf1451814194ad46ad5a6cde61cfaece83b99f3b30a8b8634
 size 891548548

run_t5.sh

@@ -21,11 +21,11 @@ mkdir -p "${MODEL_DIR}/runs"
     --learning_rate="5e-3" \
     --dtype="bfloat16" \
     --overwrite_output_dir \
-    --num_train_epochs="1" \
-    --logging_steps="20" \
+    --num_train_epochs="3" \
+    --logging_steps="50" \
     --save_steps="2000" \
     --eval_steps="10000000" \
-    --resume_from_checkpoint="${MODEL_DIR}/ckpt-14000" \
+    --resume_from_checkpoint="${MODEL_DIR}/ckpt-16000" \
     --warmup_steps="3413" \
     --push_to_hub
 

runs/Jul10_09-56-52_t1v-n-0e7426e8-w-0/events.out.tfevents.1625912272.t1v-n-0e7426e8-w-0.33788.3.v2

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:627d89553b28ce290576ab6fc5ace8c1eb487b1230497a2a37482f94c102a957
-size 44554

runs/Jul11_09-15-07_t1v-n-0e7426e8-w-0/events.out.tfevents.1625995853.t1v-n-0e7426e8-w-0.145718.3.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ecdd317adb51d2b44773888aaa52793f97b5af475a8f35560774d02bd6ae20a2
-size 300940
+oid sha256:f1aed2c3d6af21a006e34d03f0d6071dd41995d510e1cee680aad026186ca98f
+size 303949

runs/{Jul09_21-28-23_t1v-n-0e7426e8-w-0/events.out.tfevents.1625866241.t1v-n-0e7426e8-w-0.417321.3.v2 → Jul11_12-53-41_t1v-n-0e7426e8-w-0/events.out.tfevents.1626008983.t1v-n-0e7426e8-w-0.161493.3.v2}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:efed27a99c31a4b3656e323ded245a663569c2d1609cf015f02374f2897a6022
-size 31755
+oid sha256:ba46ea33a14787fe4fc81fe586551875621a7f46301e0a680643702851773220
+size 300067