Saving weights and logs of step 1000

events.out.tfevents.1642748650.t1v-n-ccbf3e94-w-0.1812628.3.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:65b2aa750afc378e6c1e0312baa298581ab407c4d4755726b5f00c2b901c6e2b
-size 13552832
+oid sha256:1fda2929c8ecb145d559f38f289ce71595c6bb5bee9a7277c7dead9fee0f92c2
+size 13702044

events.out.tfevents.1642860109.t1v-n-ccbf3e94-w-0.1929147.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c16c363c5b73e3712b3357517bdd1f1b30c385011cecdbcc4bd28fa0eafc3c7
+size 146996

events.out.tfevents.1642861064.t1v-n-ccbf3e94-w-0.1931366.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d15ef0560e48098ae925c8c8da96ea7f88d33b7468389879f03f587b776c4e8f
+size 147136

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5a8dfbba88992df82eaced28bc4267b817d61d09df4f33a3692905bc625c74b5
+oid sha256:57204ed91836522ae201803961e2c406b3c5a6ab32b5fe35434f1d5c3b0d1ec5
 size 498796983

run_128_recover_1e.sh

@@ -0,0 +1,28 @@
+python run_mlm_flax.py \
+    --output_dir="./" \
+    --model_type="roberta" \
+    --model_name_or_path="./" \
+    --config_name="roberta-base" \
+    --tokenizer_name="NbAiLab/nb-roberta-base" \
+    --dataset_name="NbAiLab/NCC" \
+    --cache_dir="/mnt/disks/flaxdisk/cache/" \
+    --max_seq_length="128" \
+    --weight_decay="0.01" \
+    --per_device_train_batch_size="232" \
+    --per_device_eval_batch_size="232" \
+    --pad_to_max_length \
+    --learning_rate="0.00031997141195461154" \
+    --warmup_steps="0" \
+    --overwrite_output_dir \
+    --num_train_epochs="9" \
+    --adam_beta1="0.9" \
+    --adam_beta2="0.98" \
+    --adam_epsilon="1e-6" \
+    --logging_steps="1000" \
+    --save_steps="1000" \
+    --eval_steps="1000" \
+    --auth_token="True" \
+    --do_train \
+    --do_eval \
+    --dtype="bfloat16" \
+    --push_to_hub