mrm8488 commited on
Commit
57af17d
1 Parent(s): 5daaa24

Update model ckpt

Browse files
events.out.tfevents.1626354152.t1v-n-cf89aecf-w-0.1016609.3.v2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0224f764bf692ac3bd6d6b158147621d510be7926328ca69c4b91c71846e0442
3
+ size 1397738
events.out.tfevents.1626364173.t1v-n-cf89aecf-w-0.1026566.3.v2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49d618c3b6ece936a9e3ae7cfcf162166289a2ad3729a48abf2dd46ac0728b02
3
+ size 40
events.out.tfevents.1626364345.t1v-n-cf89aecf-w-0.1028406.3.v2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee72a372d2f2bbe5ab9a307269cabc0e6d36909d6096ad5dbb58c546c57ad9d5
3
+ size 11059767
events.out.tfevents.1626437960.t1v-n-cf89aecf-w-0.1093227.3.v2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e852c812c8c31772be718e7995671cc231b73689899ca307c67d834e1537a51
3
+ size 14749
events.out.tfevents.1626440707.t1v-n-cf89aecf-w-0.1097198.3.v2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d91156cf42fec78270c1a1bad4c3e771c71ddc50987c79fc26604691ed4ac2bc
3
+ size 9293307
flax_model.msgpack CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a17e423adc844f2aaf44323c6738776964849df6d2ae7d0c0165bdd2267fbae1
3
  size 497764120
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5e3bd9282196a12dd5de77e69d8b231f6077e9dbe237586ad6d349fa544e281
3
  size 497764120
run.sh CHANGED
@@ -11,11 +11,12 @@
11
  --block_size="512" \
12
  --per_device_train_batch_size="64" \
13
  --per_device_eval_batch_size="64" \
14
- --learning_rate="5e-3" --warmup_steps="1000" \
15
- --adam_beta1="0.9" --adam_beta2="0.98" --weight_decay="0.01" \
 
16
  --overwrite_output_dir \
17
  --num_train_epochs="20" \
18
- --logging_steps="500" \
19
  --save_steps="2500" \
20
  --eval_steps="1000000" \
21
  --preprocessing_num_workers="64"
 
11
  --block_size="512" \
12
  --per_device_train_batch_size="64" \
13
  --per_device_eval_batch_size="64" \
14
+ --adafactor \
15
+ --learning_rate="3e-3" --warmup_steps="1000" \
16
+ --adam_beta1="0.9" --adam_beta2="0.98" --weight_decay="0.001" \
17
  --overwrite_output_dir \
18
  --num_train_epochs="20" \
19
+ --logging_steps="100" \
20
  --save_steps="2500" \
21
  --eval_steps="1000000" \
22
  --preprocessing_num_workers="64"