Saving weights and logs of step 10000

config.json

@@ -20,7 +20,7 @@
   "num_hidden_layers": 12,
   "pad_token_id": 1,
   "position_embedding_type": "absolute",
-  "torch_dtype": "bloat16",
+  "torch_dtype": "bfloat16",
   "transformers_version": "4.15.0.dev0",
   "type_vocab_size": 1,
   "use_cache": true,

events.out.tfevents.1640779990.t1v-n-358ff5d1-w-0.2119930.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce8cbb1694cdf68e6511dfe36685860589e85588faaeee1d29b3d86d2c5b5ec0
+size 1470136

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:34022cc8cc708f928034e0ae78a30f59e7249d556955b2fb7cdff3f427cb000a
+oid sha256:e126ffe6605b0ee84fefc76010690220a8469d33af635ef3ca6dcba3fe8b231f
 size 498796983

run_step1.sh

@@ -22,5 +22,5 @@
     --preprocessing_num_workers="64" \
     --auth_token="True" \
     --static_learning_rate="True" \
-    --dtype="bloat16" \
+    --dtype="bfloat16" \
     --push_to_hub