Patrick von Platen commited on
Commit
20e1eef
1 Parent(s): bf4dfc2
Files changed (2) hide show
  1. run.sh +21 -0
  2. run_mlm_flax.py +1 -0
run.sh ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ./run_mlm_flax.py \
2
+ --output_dir="./" \
3
+ --model_type="roberta" \
4
+ --config_name="${MODEL_DIR}" \
5
+ --tokenizer_name="${MODEL_DIR}" \
6
+ --dataset_name="mc4" \
7
+ --dataset_config_name="sw" \
8
+ --max_seq_length="128" \
9
+ --weight_decay="0.025" \
10
+ --per_device_train_batch_size="128" \
11
+ --per_device_eval_batch_size="128" \
12
+ --learning_rate="3e-4" \
13
+ --warmup_steps="500" \
14
+ --overwrite_output_dir \
15
+ --num_train_epochs="15" \
16
+ --adam_beta1="0.9" \
17
+ --adam_beta2="0.98" \
18
+ --logging_steps="250" \
19
+ --save_steps="1000" \
20
+ --eval_steps="1000" \
21
+ --push_to_hub
run_mlm_flax.py ADDED
@@ -0,0 +1 @@
 
1
+ /home/patrick/transformers/examples/flax/language-modeling/run_mlm_flax.py