Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
chenpangpang
transformers
Commits
e85855f2
Commit
e85855f2
authored
Dec 02, 2019
by
Lysandre
Browse files
Fix ALBERT exports with pretraining + sp classifier; Fix naming for ALBERT TF models
parent
b3d834ae
Changes
2
Show whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
20 additions
and
5 deletions
+20
-5
transformers/modeling_albert.py
transformers/modeling_albert.py
+16
-1
transformers/modeling_tf_albert.py
transformers/modeling_tf_albert.py
+4
-4
No files found.
transformers/modeling_albert.py
View file @
e85855f2
...
...
@@ -100,7 +100,22 @@ def load_tf_weights_in_albert(model, config, tf_checkpoint_path):
name
=
name
.
replace
(
"embeddings/attention"
,
"embeddings"
)
name
=
name
.
replace
(
"inner_group_"
,
"albert_layers/"
)
name
=
name
.
replace
(
"group_"
,
"albert_layer_groups/"
)
# Classifier
if
len
(
name
.
split
(
"/"
))
==
1
and
(
"output_bias"
in
name
or
"output_weights"
in
name
):
name
=
"classifier/"
+
name
# No ALBERT model currently handles the next sentence prediction task
if
"seq_relationship"
in
name
:
continue
name
=
name
.
split
(
'/'
)
# Ignore the gradients applied by the LAMB/ADAM optimizers.
if
"adam_m"
in
name
or
"adam_v"
in
name
or
"global_step"
in
name
:
logger
.
info
(
"Skipping {}"
.
format
(
"/"
.
join
(
name
)))
continue
pointer
=
model
for
m_name
in
name
:
if
re
.
fullmatch
(
r
'[A-Za-z]+_\d+'
,
m_name
):
...
...
transformers/modeling_tf_albert.py
View file @
e85855f2
...
...
@@ -31,10 +31,10 @@ import logging
logger
=
logging
.
getLogger
(
__name__
)
TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP
=
{
'albert-base-v1'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-tf_model.h5"
,
'albert-large-v1'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-tf_model.h5"
,
'albert-xlarge-v1'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-tf_model.h5"
,
'albert-xxlarge-v1'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-xxlarge-tf_model.h5"
,
'albert-base-v1'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-
v1-
tf_model.h5"
,
'albert-large-v1'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-
v1-
tf_model.h5"
,
'albert-xlarge-v1'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-
v1-
tf_model.h5"
,
'albert-xxlarge-v1'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-xxlarge-
v1-
tf_model.h5"
,
'albert-base-v2'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-base-v2-tf_model.h5"
,
'albert-large-v2'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-large-v2-tf_model.h5"
,
'albert-xlarge-v2'
:
"https://s3.amazonaws.com/models.huggingface.co/bert/albert-xlarge-v2-tf_model.h5"
,
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment