now it is working!
Browse files- __pycache__/partitions.cpython-38.pyc +0 -0
- run.sh +9 -6
- run_bak.sh → run_wiki.sh +6 -8
__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
@@ -1,7 +1,8 @@
|
|
1 |
python run_clm_mp.py \
|
2 |
-
--model_name_or_path
|
3 |
-
--tokenizer_name
|
4 |
-
--
|
|
|
5 |
--do_train \
|
6 |
--do_eval \
|
7 |
--block_size 1024 \
|
@@ -12,6 +13,8 @@ python run_clm_mp.py \
|
|
12 |
--overwrite_output_dir \
|
13 |
--output_dir /mnt/disks/flaxdisk/output/ \
|
14 |
--cache_dir /mnt/disks/flaxdisk/cache/ \
|
15 |
-
|
16 |
-
|
17 |
-
|
|
|
|
|
|
1 |
python run_clm_mp.py \
|
2 |
+
--model_name_or_path /mnt/disks/flaxdisk/norwegian-gptneo-red/ \
|
3 |
+
--tokenizer_name /mnt/disks/flaxdisk/norwegian-gptneo-red/ \
|
4 |
+
--train_file /mnt/disks/flaxdisk/corpus/social_train.json \
|
5 |
+
--validation_file /mnt/disks/flaxdisk/corpus/social_validation.json \
|
6 |
--do_train \
|
7 |
--do_eval \
|
8 |
--block_size 1024 \
|
|
|
13 |
--overwrite_output_dir \
|
14 |
--output_dir /mnt/disks/flaxdisk/output/ \
|
15 |
--cache_dir /mnt/disks/flaxdisk/cache/ \
|
16 |
+
--preprocessing_num_workers 96 \
|
17 |
+
--dtype bfloat16 \
|
18 |
+
--logging_steps 5000 \
|
19 |
+
--eval_steps 5000 \
|
20 |
+
--push_to_hub
|
run_bak.sh → run_wiki.sh
RENAMED
@@ -1,19 +1,17 @@
|
|
1 |
python run_clm_mp.py \
|
2 |
-
--model_name_or_path
|
3 |
-
--tokenizer_name
|
4 |
-
--
|
5 |
-
--validation_file /mnt/disks/flaxdisk/corpus/social_validation.json \
|
6 |
--do_train \
|
7 |
--do_eval \
|
8 |
--block_size 1024 \
|
9 |
--num_train_epochs 5 \
|
10 |
--learning_rate 4e-6 \
|
11 |
-
--per_device_train_batch_size
|
12 |
-
--per_device_eval_batch_size
|
13 |
--overwrite_output_dir \
|
14 |
--output_dir /mnt/disks/flaxdisk/output/ \
|
15 |
--cache_dir /mnt/disks/flaxdisk/cache/ \
|
16 |
--dtype bfloat16 \
|
17 |
--logging_steps 97 \
|
18 |
-
--eval_steps 96
|
19 |
-
--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 \
|
6 |
--do_eval \
|
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
|
|