You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After this PR is merged two main changes will be done:
* `onnxruntime` will be pinned to be <1.16 due some changes in
quantization recently introduced not being compatible for the moment
with the library and model.
* yolox_quantized will be downloaded from HF instead being generated at
runtime.
In order to test you can execute:
```
from unstructured_inference.inference import layout
from unstructured_inference.models.base import get_model
file = "sample-docs/example_table.jpg"
model = get_model("yolox_quantized")
doc = layout.DocumentLayout.from_image_file(
file,
model,
supplement_with_ocr_elements=True,
ocr_strategy="never",
)
```
And should show a download from HF downloading a file called
`yolox_l0.05_quantized.onnx`
0 commit comments