sumcheck and replace with new model weight

coralnet / pyspacer

Python based tools for spatial image analysis

MIT License

6 stars 2 forks source link

sumcheck and replace with new model weight #27

Closed qiminchen closed 4 years ago

qiminchen commented 4 years ago

new model weight uploaded to spacer-tools with a name of efficientnet_b0_ver1.pt
num_class changed to 1275
add sha256 sumcheck to verify downloading the weights correctly
since using the new model weight, the legacy test feature is regenerated.

qiminchen commented 4 years ago

@beijbom I uploaded the newly trained model weight to the spacer-tools as efficientnet_b0_ver1.pt and update all the places it appears in the codebase. For the sumcheck, I'm not sure if I add it correctly but pretty sure it downloads and uses the correct weight. c3dc6d304179c6729c0a0b3d4e60c728bdcf0d82687deeba54af71827467204c is generated in my local. All the test cases pass both in local and docker

beijbom commented 4 years ago

@qiminchen : please finish last items and remove "work in progress" label once ready for final review.