Upload model 1nt3s4ta (group=exp1.3-dqn_2)

README.md

@@ -31,6 +31,10 @@ and land-use grid environments.
 | **Max steps / episode** | 300 |
 | **Frame stack** | 4 |
 | **Seed** | 0 |
+| **Python (training)** | 3.11.14 |
+| **stable-baselines3** | 2.2.1 |
+| **PyTorch** | 2.10.0 |
+| **Gymnasium** | 0.29.1 |
 | **Arena size** | 60×60 cells |
 | **Camera size** | 11×11 cells |
 
@@ -56,10 +60,10 @@ The following results represent the mean performance over 500 deterministic vali
 
 ## Standalone Inference (no information-driven-uav-navigation package required)
 
-Install the minimal deps:
+Install the minimal deps (versions match the training environment):
 
 ```bash
-pip install stable-baselines3 huggingface-hub torch gymnasium
+pip install "stable-baselines3==2.2.1" "torch==2.10.0" "gymnasium==0.29.1" huggingface-hub
 ```
 
 ```python
@@ -76,10 +80,25 @@ mod  = importlib.util.module_from_spec(spec)
 sys.modules["_fe"] = mod
 spec.loader.exec_module(mod)
 
+# 1b. Stub out drone_navigation so cloudpickle can resolve ALL saved class references
+#     (lr_schedule, policy_kwargs, etc.) without the package being installed.
+import types as _types
+for _name in ["drone_navigation", "drone_navigation.models",
+              "drone_navigation.models.feature_extractor_aerial",
+              "drone_navigation.models.feature_extractor_landuse"]:
+    sys.modules.setdefault(_name, _types.ModuleType(_name))
+sys.modules["drone_navigation.models.feature_extractor_landuse"].LanduseFeaturesExtractor = mod.LanduseFeaturesExtractor
+
 # 2. Load the model -- inject the extractor class so cloudpickle can resolve it
 model = DQN.load(
     hf_hub_download(REPO_ID, "best_model.zip"),
-    custom_objects={"features_extractor_class": mod.LanduseFeaturesExtractor},
+    custom_objects={
+        "features_extractor_class": mod.LanduseFeaturesExtractor,
+        # schedule lambdas may fail to unpickle across Python versions; supply
+        # constant fallbacks (values are irrelevant for inference)
+        "lr_schedule":  lambda _: 3e-4,
+        "clip_range":   lambda _: 0.2,
+    },
     device="cpu",
 )
 

inference.py

@@ -28,6 +28,7 @@ REPO_ID = "bestak/uav-navigation-hasp"
 
 def _load_feature_extractor(repo_id: str):
     """Dynamically load LanduseFeaturesExtractor from the HF repo."""
+    import types
     from huggingface_hub import hf_hub_download
 
     fe_path = hf_hub_download(repo_id, "feature_extractor.py")
@@ -35,6 +36,21 @@ def _load_feature_extractor(repo_id: str):
     mod = importlib.util.module_from_spec(spec)
     sys.modules["_drone_nav_fe"] = mod  # register so cloudpickle can find it
     spec.loader.exec_module(mod)
+
+    # Stub out drone_navigation so cloudpickle can deserialise ALL objects stored
+    # in the zip (lr_schedule, policy_kwargs, etc.) without needing the package.
+    for _name in [
+        "drone_navigation",
+        "drone_navigation.models",
+        "drone_navigation.models.feature_extractor_aerial",
+        "drone_navigation.models.feature_extractor_landuse",
+    ]:
+        sys.modules.setdefault(_name, types.ModuleType(_name))
+    # Wire the loaded class to the original pickle path so cloudpickle resolves it.
+    sys.modules["drone_navigation.models.feature_extractor_landuse"].LanduseFeaturesExtractor = (
+        mod.LanduseFeaturesExtractor
+    )
+
     return mod.LanduseFeaturesExtractor