[TF] First changes on the road to Keras v3 #1724
Conversation
felixdittrich92
added
topic: documentation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #1724 +/- ##
==========================================
- Coverage 96.47% 96.43% -0.04%
==========================================
Files 164 164
Lines 7823 7823
==========================================
- Hits 7547 7544 -3
- Misses 276 279 +3
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
|
@odulcy-mindee I see two options:
What do you prefer ? |
|
Next steps:
|
| @@ -184,6 +184,8 @@ def main(args): | |||
|
|
|||
| # Resume weights | |||
| if isinstance(args.resume, str): | |||
| # Build the model first to load the weights | |||
| _ = model(tf.zeros((1, args.input_size, args.input_size, 3)), training=False) | |||
There was a problem hiding this comment.
Don't you need to set training=True somewhere after load_weights ?
There was a problem hiding this comment.
No that's only to build the model training=True is handled in the training scripts :)
And i hope i can remove this part with the next follow up PR where i need to implement the def build methods for most of the layers 😅
| if not params_path.is_dir() or overwrite: | ||
| with ZipFile(archive_path, "r") as f: | ||
| f.extractall(path=params_path) | ||
| # Build the model |
There was a problem hiding this comment.
@odulcy-mindee
I hope i can remove this in a follow up PR if all models have build and get_config methods where it's required
| @@ -184,6 +184,8 @@ def main(args): | |||
|
|
|||
| # Resume weights | |||
| if isinstance(args.resume, str): | |||
| # Build the model first to load the weights | |||
| _ = model(tf.zeros((1, args.input_size, args.input_size, 3)), training=False) | |||
There was a problem hiding this comment.
No that's only to build the model training=True is handled in the training scripts :)
And i hope i can remove this part with the next follow up PR where i need to implement the def build methods for most of the layers 😅
felixdittrich92
force-pushed
the
tf-h5-format
branch
from
d181043
to
d2494ed
Compare
| @@ -56,6 +56,8 @@ tf = [ | |||
| # cf. https://github.com/mindee/doctr/pull/1461 | |||
| "tensorflow>=2.11.0,<2.16.0", | |||
| "tf2onnx>=1.16.0,<2.0.0", # cf. https://github.com/onnx/tensorflow-onnx/releases/tag/v1.16.0 | |||
| # TODO: This is a temporary fix until we can upgrade to a newer version of tensorflow | |||
There was a problem hiding this comment.
Temp pin numpy until we can upgrade tensorflow to fix the issue with poetry and the failing API test (Will have effect after merging)
This PR:
is part of a series of PR's
basic elements to go further forward (only .weights.h5 and .keras (full model) - loading possible in keras v3)
change to keras imports
Finally: enable fine tuning tf models with
pretrainedalso if the vocab / classes differs !! 💯 (same we have already a long time ago added to PT - which wasn't possible with loading from/weightsin TF)all models works as expected
Next steps:
buildandget_configmethods for proper loading & saving (serialization)Any feedback is welcome 🤗
NOTE: It's still fully backwards compatible - but marking as breaking change because that will be the last release where you can load
/weights- so we have to mention this in the next release notes@odulcy-mindee models upload :)