Saving weights and logs of step 10000
Browse files- events.out.tfevents.1641904240.t1v-n-358ff5d1-w-0.938257.3.v2 +3 -0
- flax_model.msgpack +1 -1
- run_step3.sh +26 -0
- run_step4.sh +26 -0
events.out.tfevents.1641904240.t1v-n-358ff5d1-w-0.938257.3.v2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3006ce33284cc447090efc6ba97568b83e67e98497dc6486a0c192eb48afe1de
|
3 |
+
size 1470136
|
flax_model.msgpack
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 498796983
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:729c1ff63f2d39cefba876c3299619b94aeb088dfc1cf8eb4e4696b081b97e56
|
3 |
size 498796983
|
run_step3.sh
ADDED
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
./run_mlm_flax.py \
|
2 |
+
--output_dir="./" \
|
3 |
+
--model_type="roberta" \
|
4 |
+
--model_name_or_path="./" \
|
5 |
+
--config_name="./" \
|
6 |
+
--tokenizer_name="./" \
|
7 |
+
--dataset_name="NbAiLab/NCC" \
|
8 |
+
--cache_dir="/mnt/disks/flaxdisk/cache/" \
|
9 |
+
--max_seq_length="128" \
|
10 |
+
--weight_decay="0.01" \
|
11 |
+
--per_device_train_batch_size="200" \
|
12 |
+
--per_device_eval_batch_size="200" \
|
13 |
+
--learning_rate="4e-4" \
|
14 |
+
--warmup_steps="10000" \
|
15 |
+
--overwrite_output_dir \
|
16 |
+
--num_train_epochs="10" \
|
17 |
+
--adam_beta1="0.9" \
|
18 |
+
--adam_beta2="0.98" \
|
19 |
+
--adam_epsilon="1e-6" \
|
20 |
+
--logging_steps="10000" \
|
21 |
+
--save_steps="10000" \
|
22 |
+
--eval_steps="10000" \
|
23 |
+
--preprocessing_num_workers="64" \
|
24 |
+
--auth_token="True" \
|
25 |
+
--dtype="bfloat16" \
|
26 |
+
--push_to_hub
|
run_step4.sh
ADDED
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
./run_mlm_flax.py \
|
2 |
+
--output_dir="./" \
|
3 |
+
--model_type="roberta" \
|
4 |
+
--model_name_or_path="./" \
|
5 |
+
--config_name="./" \
|
6 |
+
--tokenizer_name="./" \
|
7 |
+
--dataset_name="NbAiLab/NCC" \
|
8 |
+
--cache_dir="/mnt/disks/flaxdisk/cache/" \
|
9 |
+
--max_seq_length="128" \
|
10 |
+
--weight_decay="0.01" \
|
11 |
+
--per_device_train_batch_size="200" \
|
12 |
+
--per_device_eval_batch_size="200" \
|
13 |
+
--learning_rate="2e-4" \
|
14 |
+
--warmup_steps="10000" \
|
15 |
+
--overwrite_output_dir \
|
16 |
+
--num_train_epochs="10" \
|
17 |
+
--adam_beta1="0.9" \
|
18 |
+
--adam_beta2="0.98" \
|
19 |
+
--adam_epsilon="1e-6" \
|
20 |
+
--logging_steps="10000" \
|
21 |
+
--save_steps="10000" \
|
22 |
+
--eval_steps="10000" \
|
23 |
+
--preprocessing_num_workers="64" \
|
24 |
+
--auth_token="True" \
|
25 |
+
--dtype="bfloat16" \
|
26 |
+
--push_to_hub
|