In 5c, run_train() creates a randomly initialized model and immediately saves it as the initial self-play checkpoint. base.load_model_path is never applied.
Because of this, the self-play pool is bootstrapped with random weights rather than the requested checkpoint and training also starts from those random weights without warning.
Problematic code: src/pufferl.cu:3240
Expected: load base.load_model_path and synchronize async actor weights before saving or registering the initial checkpoint.
In
5c,run_train()creates a randomly initialized model and immediately saves it as the initial self-play checkpoint.base.load_model_pathis never applied.Because of this, the self-play pool is bootstrapped with random weights rather than the requested checkpoint and training also starts from those random weights without warning.
Problematic code: src/pufferl.cu:3240
Expected: load
base.load_model_pathand synchronize async actor weights before saving or registering the initial checkpoint.