gt-ripl / featmatch Goto Github PK

Hi, Can you please highlight the part of the code that performs backpropagation, and updation of the model (via loss.backward(), opt.step() ) ?

Hi,

Thanks for the awesome work! I wonder, what kind of information is stored in this file zca file and what is the purpose of doing this? Many thanks!
UPDATE: [Resolved]

Hi, thank you for the great paper! Is there a specific way to set the following hyperparameters? As I plan to implement FeatMatch to other datasets

"pk": 20,
"l_ratio": 0.5,
"pretrain_iters": 3000,
"cycle_iters": 75000,
"end_iters": 30000,
"coeff_rampup": 75000,
 "update_interval": 250,
 "sample_interval": 500