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

trocr viarant for Predictor runtime: no model size for input_shape…

Browse source 
Because transformers v4 and v5 API for image preprocessor differs,
and the model-internal image input sizes are actually irrelevant,
because the preprocessor will resize them anyway, and there is no
batch dimension (because the input images will have different shapes),
do not advertise this information in `.input_shape`.
This commit is contained in:
Robert Sachunsky 2026-06-03 20:51:56 +02:00
parent f447a9f248
commit 4e7e1c06b9
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/eynollah/model_zoo/model_zoo.py
View file

@ -379,9 +379,9 @@ class EynollahModelZoo:
return text, conf return text, conf
model.predict_on_batch = predict_torch model.predict_on_batch = predict_torch
# not actually needed (image processor does resize itself) # not actually needed (image processor does resize itself)
# no batch dimension (images passed as list w/ varying shapes)
model.input_shape = (None, model.input_shape = (None,
proc.image_processor.size.height, None,
proc.image_processor.size.width,
len(proc.image_processor.image_mean)) len(proc.image_processor.image_mean))
return model return model