Update models/GroundingDINO/ms_deform_attn.py

Added absolute import to top

Forced import _C

Update ms_deform_attn.py in main/models/GroundingDINO/ with different MSDA import

Update build_ops.sh

Printing CUDA_HOME location

removed export CUDA_HOME

Set CUDA_HOME in build_ops.sh

Running build_ops.sh directly instead of relying on wheel

Update build_ops.sh

Create test.py

Removed call to check_ms_deform

Changed running test code inline to running it in separate script

Added the functions import after installing msdeformable attention

Moved __future__ imports to the top

Added a check to check the install of MS Deform Atten

Update models/GroundingDINO/ms_deform_attn.py

printing on deformable self attention

Changed checkpointing in config

Removed toplevel checkpoint setting for transformer

set model.transformer not model. for checkpointing

Turned off use_checkpoint and use_transformer_ckpt for model

More debug printing inside transformer encoder

Added print statements in transformer file for debugging

More debugging printing

Printing layer index while debugging

More printing for debugging purposes

More debugging print statements

Set duration parameter to 120 s on tutorial counting function decorator

Removed print statements from model file

Made call of model to device global in app.py file

Added debugging texts to inside of gdino module

Moved model to gpu before calling count functions

Update app.py

printing devices

Changed how device being allocated

Tried commenting out model.cpu() part of tutorial count function

Removed GPU for building models and transforms

Returned model to cpu after the execution of each gpu function

Added @spaces.gpu decorator and switched to gpu officially

Changed py310 to cp310 in multiscale deformable attention file name

Rename MultiScaleDeformableAttention-1.0-py310-cp310-linux_x86_64.whl to MultiScaleDeformableAttention-1.0-cp310-cp310-linux_x86_64.whl

Added ushlex

Added import shlex

Changed to direct call to subprocess.run

Added call to build_multiscale_deform

Added build_multiscale_deform_atten with new python 3.10 wheel file

Create build_multiscale_deform.sh

Upload MultiScaleDeformableAttention-1.0-py310-cp310-linux_x86_64.whl

Quick test