backup before making experiment
Browse files- __pycache__/partitions.cpython-38.pyc +0 -0
- run.sh +4 -5
__pycache__/partitions.cpython-38.pyc
CHANGED
Binary files a/__pycache__/partitions.cpython-38.pyc and b/__pycache__/partitions.cpython-38.pyc differ
|
|
run.sh
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
python run_clm_mp.py \
|
2 |
-
--model_name_or_path
|
3 |
--tokenizer_name gpt2 \
|
4 |
--dataset_name wikitext --dataset_config_name wikitext-2-raw-v1 \
|
5 |
--do_train \
|
@@ -7,12 +7,11 @@ python run_clm_mp.py \
|
|
7 |
--block_size 1024 \
|
8 |
--num_train_epochs 5 \
|
9 |
--learning_rate 4e-6 \
|
10 |
-
--per_device_train_batch_size
|
11 |
-
--per_device_eval_batch_size
|
12 |
--overwrite_output_dir \
|
13 |
--output_dir /mnt/disks/flaxdisk/output/ \
|
14 |
--cache_dir /mnt/disks/flaxdisk/cache/ \
|
15 |
--dtype bfloat16 \
|
16 |
--logging_steps 97 \
|
17 |
-
--eval_steps 96
|
18 |
-
--push_to_hub
|
|
|
1 |
python run_clm_mp.py \
|
2 |
+
--model_name_or_path "." \
|
3 |
--tokenizer_name gpt2 \
|
4 |
--dataset_name wikitext --dataset_config_name wikitext-2-raw-v1 \
|
5 |
--do_train \
|
|
|
7 |
--block_size 1024 \
|
8 |
--num_train_epochs 5 \
|
9 |
--learning_rate 4e-6 \
|
10 |
+
--per_device_train_batch_size 2 \
|
11 |
+
--per_device_eval_batch_size 2 \
|
12 |
--overwrite_output_dir \
|
13 |
--output_dir /mnt/disks/flaxdisk/output/ \
|
14 |
--cache_dir /mnt/disks/flaxdisk/cache/ \
|
15 |
--dtype bfloat16 \
|
16 |
--logging_steps 97 \
|
17 |
+
--eval_steps 96
|
|