Saving weights and logs of step 50001
Browse files
events.out.tfevents.1629033809.t1v-n-358ff5d1-w-0.179703.3.v2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ba4cc1cf8e9c0c94ad287510f6528a228e55428935a0d1a42a7f44c5657175a7
|
3 |
+
size 40
|
events.out.tfevents.1629051212.t1v-n-358ff5d1-w-0.196309.3.v2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f9d08eae93640149e2e3aecf5727e0e10b8f93382d823a664bcad666fca39707
|
3 |
+
size 7417652
|
flax_model.msgpack
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4aaaada5492a49ba02351ecbf13b1f9789e70fe6917cf95119aa3242efe2cac0
|
3 |
+
size 498796983
|
run_stream.sh
CHANGED
@@ -6,8 +6,8 @@
|
|
6 |
--dataset_name="pere/norwegian_colossal_corpus_v2_short100k" \
|
7 |
--max_seq_length="128" \
|
8 |
--weight_decay="0.01" \
|
9 |
-
--per_device_train_batch_size="
|
10 |
-
--per_device_eval_batch_size="
|
11 |
--learning_rate="6e-4" \
|
12 |
--warmup_steps="50000" \
|
13 |
--overwrite_output_dir \
|
|
|
6 |
--dataset_name="pere/norwegian_colossal_corpus_v2_short100k" \
|
7 |
--max_seq_length="128" \
|
8 |
--weight_decay="0.01" \
|
9 |
+
--per_device_train_batch_size="232" \
|
10 |
+
--per_device_eval_batch_size="232" \
|
11 |
--learning_rate="6e-4" \
|
12 |
--warmup_steps="50000" \
|
13 |
--overwrite_output_dir \
|