Saving weights and logs of step 500

events.out.tfevents.1631216744.t1v-n-1a0a7c50-w-0.1083508.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c451c48c3b7b5fef3c9c190f79ea3daaee814784edf7dd87901b38d6ebfe076
+size 40

events.out.tfevents.1631217826.t1v-n-1a0a7c50-w-0.1102830.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbc8e1adb6803ae07acbae62936b11724dec25c7fd4e1e4336a3cbe65945f281
+size 40

events.out.tfevents.1631270664.t1v-n-1a0a7c50-w-0.1438451.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1daad6304a496c40d0f25a275a2d7d4c38b01d62204b6b95a10c56767d7d5d1
+size 40

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c694052f126d1176eaa04c0dffd77b0c49a301f6677b83a9d39d4a61f3c59ccc
+oid sha256:93dddc3e0a75161cc9c8ca31b5175a6faf380aa20a4bab6eef572ec3f4dd692c
 size 5262371934

run.sh

@@ -6,14 +6,15 @@ python run_clm_mp.py \
      	--do_train \
        	--do_eval  \
      	--block_size 1024 \
-   	--num_train_epochs 10 \
+   	--num_train_epochs 5 \
    	--learning_rate 4e-6 \
-	--per_device_train_batch_size 3 \
-       	--per_device_eval_batch_size 3 \
+	--per_device_train_batch_size 2 \
+       	--per_device_eval_batch_size 2 \
     	--overwrite_output_dir \
-	--output_dir /mnt/disks/flaxdisk/norwegian-gptneo-blue \
+	--output_dir "./" \
 	--cache_dir /mnt/disks/flaxdisk/cache/ \
-       	--dtype bfloat16 \
-   	--logging_steps 97 \
-       	--eval_steps 96\
+	--preprocessing_num_workers 96 \
+	--dtype bfloat16 \
+   	--logging_steps 5000 \
+       	--eval_steps 5000 \
 	--push_to_hub