easonabc-public
/
paia-hw3
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

print acc reward

This commit is contained in:
Eason 2024-04-26 13:34:20 +08:00
parent b4840596cb
commit 833fcb8978
2 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
justfile
View File

@ -5,6 +5,7 @@ build:
build-cuda: build-cuda:
cd pyr && cargo build --release --features cuda cd pyr && cargo build --release --features cuda
train level: train level:
python -m mlgame -f 400 -i ./ml/ml_play_pyr_test.py -i ./ml/ml_play_manual_2P.py . --sound off --level {{level}} --game_times 3 | grep Rewards
run level: run level:
python -m mlgame -f 400 -i ./ml/ml_play_pyr_test.py -i ./ml/ml_play_manual_2P.py . --sound off --level {{level}} --game_times 3 python -m mlgame -f 400 -i ./ml/ml_play_pyr_test.py -i ./ml/ml_play_manual_2P.py . --sound off --level {{level}} --game_times 3
clean: clean:

3
pyr/src/app/agent.rs
View File

@ -186,7 +186,6 @@ impl AIAgent {
// Compare the estimated rewards with the maximum expected rewards and // Compare the estimated rewards with the maximum expected rewards and
// perform the backward step. // perform the backward step.
let loss = huber_loss(1.0_f32)(&x, &y); let loss = huber_loss(1.0_f32)(&x, &y);
log::trace!("loss: {:?}", loss);
self.optimizer self.optimizer
.backward_step(&Tensor::new(&[loss], &self.device).unwrap()) .backward_step(&Tensor::new(&[loss], &self.device).unwrap())
.unwrap(); .unwrap();
@ -195,6 +194,8 @@ impl AIAgent {
self.memory.clear(); self.memory.clear();
if CONFIG.train { if CONFIG.train {
self.var_map.save("model.bin").unwrap(); self.var_map.save("model.bin").unwrap();
log::info!("Rewards {}", self.accumulate_rewards as i64);
self.accumulate_rewards = 0.0;
log::info!("model.bin saved!"); log::info!("model.bin saved!");
} }
} }