wip

bot/src/burnrl/dqn_model.rs

@@ -2,6 +2,7 @@ use crate::burnrl::utils::soft_update_linear;
 use burn::module::Module;
 use burn::nn::{Linear, LinearConfig};
 use burn::optim::AdamWConfig;
+use burn::record::{CompactRecorder, Recorder};
 use burn::tensor::activation::relu;
 use burn::tensor::backend::{AutodiffBackend, Backend};
 use burn::tensor::Tensor;
@@ -138,5 +139,15 @@ pub fn run<E: Environment, B: AutodiffBackend>(
         }
     }
 
+    // Save
+    let path = "models/burn_dqn".to_string();
+    let inference_network = agent.model().clone().into_record();
+    let recorder = CompactRecorder::new();
+    let model_path = format!("{}_model.burn", path);
+    println!("Modèle sauvegardé : {}", model_path);
+    recorder
+        .record(inference_network, model_path.into())
+        .unwrap();
+
     agent.valid()
 }

bot/src/burnrl/main.rs

@@ -1,16 +1,13 @@
+use bot::burnrl::{dqn_model, environment, utils::demo_model};
 use burn::backend::{Autodiff, NdArray};
 use burn_rl::base::ElemType;
-use bot::burnrl::{
-    dqn_model,
-    environment,
-    utils::demo_model,
-};
 
 type Backend = Autodiff<NdArray<ElemType>>;
 type Env = environment::TrictracEnvironment;
 
 fn main() {
-    let agent = dqn_model::run::<Env, Backend>(512, false); //true);
+    let num_episodes = 3;
+    let agent = dqn_model::run::<Env, Backend>(num_episodes, false); //true);
 
     demo_model::<Env>(agent);
 }

bot/src/burnrl/utils.rs

@@ -12,6 +12,7 @@ pub fn demo_model<E: Environment>(agent: impl Agent<E>) {
         if let Some(action) = agent.react(&state) {
             let snapshot = env.step(action);
             state = *snapshot.state();
+            // println!("{:?}", state);
             done = snapshot.done();
         }
     }

doc/backlog.md

@@ -1,11 +1,17 @@
 # Backlog
 
-position dans tutoriel :
-
 ## DONE
 
 ## TODO
 
+- bot burn
+  - train = `just trainbot`
+    - durée d'entrainement selon params ?
+  - save
+  - load and run against default bot
+  - many configs, save models selon config
+  - retrain against himself ?
+
 ### Doc
 
 Cheatsheet : arbre des situations et priorité des règles

justfile

@@ -21,4 +21,5 @@ trainbot:
   #python ./store/python/trainModel.py
   # cargo run --bin=train_dqn # ok
   # cargo run --bin=train_burn_rl # doesn't save model
-  cargo run --bin=train_dqn_full
+  # cargo run --bin=train_dqn_full
+  cargo run --bin=train_dqn_burn