Saving weights and logs of step 500

config.json

@@ -65,7 +65,7 @@
     }
   },
   "tokenizer_class": "GPT2Tokenizer",
-  "transformers_version": "4.9.0.dev0",
+  "transformers_version": "4.11.0.dev0",
   "use_cache": true,
   "vocab_size": 50264,
   "window_size": 256

events.out.tfevents.1631191833.t1v-n-4e27a527-w-0.1937094.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:121d2e96899f6772a57b1228900bcbfd6ba74fa50e4982f41b305cb7571b56c2
+size 40

flax_model.msgpack

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c694052f126d1176eaa04c0dffd77b0c49a301f6677b83a9d39d4a61f3c59ccc
+oid sha256:21114569ff0ec5434ff6b47f5b8b254ab174c5ba6c4ece5f7edd4b8a7526ece0
 size 5262371934

run.sh

@@ -11,7 +11,7 @@ python run_clm_mp.py \
 	--per_device_train_batch_size 2 \
        	--per_device_eval_batch_size 2 \
     	--overwrite_output_dir \
-	--output_dir /mnt/disks/flaxdisk/output/ \
+	--output_dir "./" \
 	--cache_dir /mnt/disks/flaxdisk/cache/ \
 	--preprocessing_num_workers 96 \
 	--dtype bfloat16 \