Open MichaelHoltonPrice opened 10 months ago
yeah also getting this same issue. If I change the key to "default", I'm getting shape error mismatches between the expected model config and the checkpoint.
EDIT: Ok, seems that the EdgeSAM repo contains the files needed to get this running: https://github.com/chongzhou96/EdgeSAM/tree/master
Namely, build_sam.py, common.py and rep_vit.py
I copied these over to the relevant locations in my autodistill lib (highly unrecommended) and the runs fine with "edge_sam" as the key.
Odd that the dependencies for this EdgeSAM weren't updated/modified with this repo
hello ran into the same error while running : when running import sys sys.path.append("..") from segment_anything import sam_model_registry, SamPredictor
sam_checkpoint = "/content/EfficientSAM/edge_sam.pth" model_type = "edge_sam"
device = "cuda"
sam = sam_model_registrymodel_type sam.to(device=device)
predictor = SamPredictor(sam) --------------------------------------------------------------------------- KeyError Traceback (most recent call last) /tmp/ipykernel_11959/398106226.py in <cell line: 10>() 8 device = "cuda" 9 ---> 10 sam = sam_model_registrymodel_type 11 sam.to(device=device) 12
KeyError: 'edge_sam is the solution already found ?
I receive the following error.
base_model = GroundedEdgeSAM( ... ontology=CaptionOntology( ... { ... "person": "person", ... "forklift": "forklift", ... } ... ) ... ) Loading EdgeSAM... Traceback (most recent call last): File "", line 1, in
File "C:\Users\mpatm\AppData\Local\Programs\Python\Python310\lib\site-packages\autodistill_grounded_edgesam\grounded_edgesam_model.py", line 77, in init
self.predictor = check_dependencies()
File "C:\Users\mpatm\AppData\Local\Programs\Python\Python310\lib\site-packages\autodistill_grounded_edgesam\grounded_edgesam_model.py", line 59, in check_dependencies
sam = sam_model_registrySAM_ENCODER_VERSION.to(
KeyError: 'edge_sam'
I've confirmed that, indeed, the sam_model_registry dictionary does not contain a key named edge_sam. This code ran without error the first time I ran it, but not the second:
https://github.com/autodistill/autodistill-grounded-edgesam?ref=blog.roboflow.com
I worry that the automatic installs in grounded_edgesam_model.py messed up my configuration. It seems like a bad idea to embed clones and installs in that script. It would be better to trust your users to do the appropriate installs, at most raising an error if there is an issue; this is in line with pythonic principles.