qurator-spk/eynollah
1
0
Fork 0
mirror of https://github.com/qurator-spk/eynollah.git synced 2026-06-16 09:59:13 +02:00
Code Issues Projects Releases Wiki Activity

ModelZoo TF-Serving backend: deal with buggy .inputs signature…

Browse source 
work around TF bug that adds captured/unknown inputs to function signature
This commit is contained in:
Robert Sachunsky 2026-06-12 14:56:44 +02:00
parent 45c92eada2
commit 94082bc64a
1 changed files with 13 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

14
src/eynollah/model_zoo/model_zoo.py
View file

@ -281,7 +281,19 @@ class EynollahModelZoo:
self._configure_tf_device(model_category, device=device) self._configure_tf_device(model_category, device=device)
model = tf.saved_model.load(model_path) model = tf.saved_model.load(model_path)
model.predict_on_batch = model.serve model.predict_on_batch = model.serve
model.input_shape = tuple(model.signatures.get('serving_default').inputs[0].shape) spec = model.signatures['serving_default']
# some models receive lots of additional/internal
# (unknown) captured inputs polluting .inputs
# TF>=2.16 has spec.function_type.flat_inputs
# this non-public API works:
# input_spec = spec.inputs[:len(spec._arg_keywords)]
# but perhaps this is most reliable:
input_spec = tf.nest.flatten(spec.structured_input_signature, True)
input_spec = [tuple(i.shape) for i in input_spec]
if len(input_spec) > 1:
model.input_shape = tuple(input_spec)
else:
model.input_shape = input_spec[0]
return model return model