Request for PathoSAM histopathology checkpoint weights (vit_b_histopathology.pth)

[benbettitwassila05-cell]

Hello,

I am using the PathoSAM codebase and installed all dependencies successfully. I found in checkpoints/ directory a file named sam_vit_b.pth. However, when attempting to run the segmentation script, I get the error:

RuntimeError: You have passed 'amg=False', but your model does not contain a segmentation decoder.

It seems that the checkpoint I have is not a proper PathoSAM histopathology model (it lacks the segmentation decoder).

Could you please share the correct checkpoint for vit_b_histopathology, or the appropriate model weights for automatic segmentation?

Thank you very much for your valuable work and support

[anwai98]

Hi @benbettitwassila05-cell,

Thanks for your interest in patho-sam.

Could you please share me your scripts on how are you loading thepatho-sam weights?

PS. patho-sam (with micro-sam in the background) automates the entire download process for you.

[benbettitwassila05-cell]

Hello, Thank you for your reply. Here is the command/script I used to run PathoSAM: python -m patho_sam.automatic_segmentation \ -i C:\Users\LENOVO\Desktop\PFE\Histo_Thyro\images\tsg10_1.JPG \ -o C:\Users\LENOVO\Desktop\PFE\Histo_Thyro\results\tsg10_1_mask.tif \ -m vit_b_histopathology \ -c C:\Users\LENOVO\Desktop\PFE\Histo_Thyro\patho-sam-master\checkpoints\sam_vit_b.pth Le mer. 1 oct. 2025 à 11:17, Anwai Archit ***@***.***> a écrit :

…

*anwai98* left a comment (computational-cell-analytics/patho-sam#51) <#51 (comment)> Hi @benbettitwassila05-cell <https://github.com/benbettitwassila05-cell>, Thanks for your interest in patho-sam. Could you please share me your scripts on how are you loading thepatho-sam weights? PS. patho-sam (with micro-sam in the background) automates the entire download process for you. — Reply to this email directly, view it on GitHub <#51 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BYEPVXHHE4KJG5XLLG3YEYL3VOLY3AVCNFSM6AAAAACH7NALCOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGNJVGQ2DQNJYHE> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[anwai98]

Hi @benbettitwassila05-cell,

Could you please tell me where did you get this model weights from? (~sam_vit_b.pth) Is it a model you trained yourself on your data, or is it one of our models? And would you like to use our patho-sam models or your own model?

[benbettitwassila05-cell]

Hello, thank you for your reply. The file `sam_vit_b.pth` was not trained by me. I found it in the `checkpoints/` directory after downloading the Patho-SAM repository from GitHub: https://github.com/computational-cell-analytics/patho-sam I would like to use your official Patho-SAM histopathology models. Le mer. 1 oct. 2025 à 12:30, Anwai Archit ***@***.***> a écrit :

…

*anwai98* left a comment (computational-cell-analytics/patho-sam#51) <#51 (comment)> Hi @benbettitwassila05-cell <https://github.com/benbettitwassila05-cell>, Could you please tell me where did you get this model weights from? (~ sam_vit_b.pth) Is it a model you trained yourself on your data, or is it one of our models? And would you like to use our patho-sam models or your own model? — Reply to this email directly, view it on GitHub <#51 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BYEPVXAAWVID32ZFNJA7FAL3VOUKZAVCNFSM6AAAAACH7NALCOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGNJVG4YDGNJWGU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[anwai98]

Thanks for the response!

For using our patho-sam generalist models, you don't need any checkpoints paths to be provided to the inference setup.

Just use the command you provided without any -c argument (should look something like below):

python -m patho_sam.automatic_segmentation \
    -i C:\Users\LENOVO\Desktop\PFE\Histo_Thyro\images\tsg10_1.JPG \
    -o C:\Users\LENOVO\Desktop\PFE\Histo_Thyro\results\tsg10_1_mask.tif \
    -m vit_b_histopathology

That should do the trick! (and make sure your terminal / system has access to the internet so that it can download the model)

PS. I'll keep the issue open for now. Let us know if this works for you!