Saving weights and logs of epoch 1

events.out.tfevents.1626345259.t1v-n-b95d739e-w-0.658394.3.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d513f0a62c7e17cf589d8d1a18ba90a33032ef92e9d9b82d0ae97293b6b045e
+size 78623

flax_model.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba730ec436fcc4bf257bfe0294d8c3a73ebdf7d5d4d28940fe865d60017e05ca
+size 497764120

nohup.out

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e567ef92b569fa9ec4fae1bf321ea00722baa03cc83002a3b356f7c1752de84a
+size 147736

run.sh

@@ -1,6 +1,5 @@
 #!/usr/bin/env bash
 python3 run_clm_flax.py \
-    --model_name_or_path="flax_model.msgpack" \
     --output_dir="./" \
     --model_type="gpt2" \
     --config_name="./" \

run_clm_flax.py

@@ -388,10 +388,10 @@ def main():
         cc100 = cc100.remove_columns("id")
 
         # combine datasets
-        datasets = concatenate_datasets([oscar, mc4, cc100, wiki["train"]])
+        dataset = concatenate_datasets([oscar, mc4, cc100, wiki["train"]])
         # split train and validation
         # note: renamed `validation` key to `test` everywhere else in the script
-        datasets = datasets.train_test_split(
+        dataset = dataset.train_test_split(
             test_size=data_args.validation_split_percentage / 100, seed=42
         )