Saving weights and logs of step 2500

events.out.tfevents.1631271978.t1v-n-d4d6e0cd-w-0.40355.0.v2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ebdf2e01eb26c6600a5107d32f763353b0a79d7a3a08c3ab08e591af42ec8531
-size 10752297
+oid sha256:d07c587c9b4cadb323b4f6475cecfeb801baee195091d11f12d6098e3006aff3
+size 10826867

events.out.tfevents.1631342182.t1v-n-d4d6e0cd-w-0.113370.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff8f9e2af2235985b80b7089527b58de5163da05c658525e8d456081e301976e
+size 40

events.out.tfevents.1631344615.t1v-n-d4d6e0cd-w-0.3764.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1b79a991f6d34452ce746fa2d94ed14f45212826894b0378c4c5988f335928b
+size 40

events.out.tfevents.1631345739.t1v-n-d4d6e0cd-w-0.5509.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0987808e42cb735cf737b686f88f56b8c1bd9d42e75a574b00103ab4b15011b
+size 367912

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f22233ccf4a61962d4a8e18c37e86bb8dacf6e687ba8cf085635bbdf310e69d8
+oid sha256:f1ecbffd7b0c9c000e97760acd00710cdc817bbbb4328e36499aab3087bb4366
 size 1100762015

run.sh

@@ -7,8 +7,8 @@
     --validation_file /mnt/disks/flaxdisk/NCC_nb_nn_balanced/nb_nn_balanced_shuffled_100k_validation.json \
     --max_seq_length="512" \
     --weight_decay="0.01" \
-    --per_device_train_batch_size="32" \
-    --per_device_eval_batch_size="32" \
+    --per_device_train_batch_size="16" \
+    --per_device_eval_batch_size="16" \
     --learning_rate="8e-3" \
     --warmup_steps="2000" \
     --overwrite_output_dir \

run_recover_1e.sh

@@ -0,0 +1,26 @@
+./run_t5_mlm_flax.py \
+    --output_dir="./" \
+    --model_type="t5" \
+    --model_name_or_path="." \
+    --config_name="./" \
+    --tokenizer_name="./" \
+    --train_file /mnt/disks/flaxdisk/NCC_nb_nn_balanced/nb_nn_balanced_shuffled_100k_train.json \
+    --validation_file /mnt/disks/flaxdisk/NCC_nb_nn_balanced/nb_nn_balanced_shuffled_100k_validation.json \
+    --max_seq_length="512" \
+    --weight_decay="0.01" \
+    --per_device_train_batch_size="32" \
+    --per_device_eval_batch_size="32" \
+    --learning_rate="0.006964816711843014" \
+    --warmup_steps="0" \
+    --overwrite_output_dir \
+    --cache_dir /mnt/disks/flaxdisk/cache/ \
+    --num_train_epochs="14" \
+    --adam_beta1="0.9" \
+    --adam_beta2="0.98" \
+    --logging_steps="500" \
+    --save_steps="2500" \
+    --eval_steps="2500" \
+    --preprocessing_num_workers 96 \
+    --adafactor \
+    --push_to_hub
+