Upload folder using huggingface_hub

.summary/0/events.out.tfevents.1689539907.qgallouedec-MS-7C84

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11dbd45361a7306781851e4f70827976a9112447d6a551f1f3c47078ff98b81d
+size 582049

README.md

@@ -15,7 +15,7 @@ model-index:
       type: door-unlock-v2
     metrics:
     - type: mean_reward
-      value: 805.13 +/- 11.00
+      value: 805.08 +/- 13.10
       name: mean_reward
       verified: false
 ---

checkpoint_p0/best_000014776_7565312_reward_804.786.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00b92ede5c8bcf47afa74f4697412829565b02873e8ac83846e873fc4193f20b
+size 98239

checkpoint_p0/checkpoint_000019232_9846784.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77425cc817d7cb8661b9a85f13aaf6157f4fa8a7643c83ca112b32faf83eaf09
+size 98567

checkpoint_p0/checkpoint_000019544_10006528.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cb666802137fed58f5f759fcb8f008202e2d0bb927f3baeb683117ac74d9028a
+oid sha256:f8299ced3eb43fb019a9be6f77d9d74c77e91bfd8f25c35be70c038cdc2fe55e
 size 98567

config.json

@@ -128,7 +128,7 @@
     "wandb_user": "qgallouedec",
     "wandb_project": "sample_facotry_metaworld"
   },
-  "git_hash": "66db1b7a27030aa65fcfa2d6e3503089a7cff207",
+  "git_hash": "0010a8dda1f21d81bec67d77b900206dea2683b2",
   "git_repo_name": "https://github.com/huggingface/gia",
-  "wandb_unique_id": "door-unlock-v2_20230708_172532_864739"
+  "wandb_unique_id": "door-unlock-v2_20230716_223825_780327"
 }

git.diff

@@ -1,3 +1,16 @@
+diff --git a/data/envs/metaworld/train_all.sh b/data/envs/metaworld/train_all.sh
+index dbf328a..c393191 100755
+--- a/data/envs/metaworld/train_all.sh
++++ b/data/envs/metaworld/train_all.sh
+@@ -4,7 +4,7 @@ ENVS=(
+     assembly
+     basketball
+     bin-picking
+-    box-close
++    #box-close
+     button-press-topdown
+     button-press-topdown-wall
+     button-press
 diff --git a/gia/eval/callback.py b/gia/eval/callback.py
 index 5c3a080..4b6198f 100644
 --- a/gia/eval/callback.py
@@ -14,45 +27,21 @@ index 5c3a080..4b6198f 100644
  from gia.config import Arguments
  from gia.eval.utils import is_slurm_available
  
-diff --git a/gia/eval/evaluator.py b/gia/eval/evaluator.py
-index 91b645c..3e2cae7 100644
---- a/gia/eval/evaluator.py
-+++ b/gia/eval/evaluator.py
-@@ -1,3 +1,5 @@
-+from typing import Optional
-+
- import torch
- 
- from gia.config.arguments import Arguments
-@@ -5,11 +7,12 @@ from gia.model import GiaModel
- 
- 
- class Evaluator:
--    def __init__(self, args: Arguments, task: str) -> None:
-+    def __init__(self, args: Arguments, task: str, mean_random: Optional[float] = None) -> None:
-         self.args = args
-         self.task = task
-+        self.mean_random = mean_random
- 
--    @torch.no_grad()
-+    @torch.inference_mode()
-     def evaluate(self, model: GiaModel) -> float:
-         return self._evaluate(model)
- 
 diff --git a/gia/eval/rl/envs/core.py b/gia/eval/rl/envs/core.py
-index ec5e5b2..eeaf7cb 100644
+index ec5e5b2..3294471 100644
 --- a/gia/eval/rl/envs/core.py
 +++ b/gia/eval/rl/envs/core.py
-@@ -177,7 +177,6 @@ def make(task_name: str, num_envs: int = 1):
- 
-     elif task_name.startswith("metaworld"):
-         import gymnasium as gym
--        import metaworld
+@@ -180,7 +180,7 @@ def make(task_name: str, num_envs: int = 1):
+         import metaworld
  
          env_id = TASK_TO_ENV_MAPPING[task_name]
-         env = gym.vector.SyncVectorEnv([lambda: gym.make(env_id)] * num_envs)
+-        env = gym.vector.SyncVectorEnv([lambda: gym.make(env_id)] * num_envs)
++        env = gym.make(env_id)
+ 
+     else:
+         raise ValueError(f"Unknown task name: {task_name}")
 diff --git a/gia/eval/rl/gia_agent.py b/gia/eval/rl/gia_agent.py
-index f0d0b9b..39dc0d2 100644
+index f0d0b9b..255beda 100644
 --- a/gia/eval/rl/gia_agent.py
 +++ b/gia/eval/rl/gia_agent.py
 @@ -54,7 +54,7 @@ class GiaAgent:
@@ -60,62 +49,19 @@ index f0d0b9b..39dc0d2 100644
          self.deterministic = deterministic
          self.device = next(model.parameters()).device
 -        self._max_length = self.model.config.max_position_embeddings - 10
-+        self._max_length = self.model.config.max_position_embeddings - 100  # TODO: fix this
++        self._max_length = self.model.config.max_position_embeddings - 100
  
          if isinstance(observation_space, spaces.Box):
              self._observation_key = "continuous_observations"
-@@ -75,6 +75,11 @@ class GiaAgent:
-     ) -> Tuple[Tuple[Tensor, Tensor], ...]:
-         return tuple((k[:, :, -self._max_length :], v[:, :, -self._max_length :]) for (k, v) in past_key_values)
- 
-+    def set_model(self, model: GiaModel) -> None:
-+        self.model = model
-+        self.device = next(model.parameters()).device
-+        self._max_length = self.model.config.max_position_embeddings
-+
-     def reset(self, num_envs: int = 1) -> None:
-         if self.prompter is not None:
-             prompts = self.prompter.generate_prompts(num_envs)
 diff --git a/gia/eval/rl/gym_evaluator.py b/gia/eval/rl/gym_evaluator.py
-index f8531ee..754c05d 100644
+index f8531ee..71e0fdc 100644
 --- a/gia/eval/rl/gym_evaluator.py
 +++ b/gia/eval/rl/gym_evaluator.py
-@@ -1,7 +1,7 @@
+@@ -1,7 +1,6 @@
  import gym
  from gym.vector.vector_env import VectorEnv
  
 -from gia.eval.mappings import TASK_TO_ENV_MAPPING
-+# from gia.eval.rl.envs.mappings import TASK_TO_ENV_MAPPING
  from gia.eval.rl.rl_evaluator import RLEvaluator
  
  
-diff --git a/gia/eval/rl/rl_evaluator.py b/gia/eval/rl/rl_evaluator.py
-index c5cc423..91189f3 100644
---- a/gia/eval/rl/rl_evaluator.py
-+++ b/gia/eval/rl/rl_evaluator.py
-@@ -8,6 +8,10 @@ from gia.eval.rl.gia_agent import GiaAgent
- 
- 
- class RLEvaluator(Evaluator):
-+    def __init__(self, args, task):
-+        super().__init__(args, task)
-+        self.agent = GiaAgent()
-+
-     def _build_env(self) -> VectorEnv:  # TODO: maybe just a gym.Env ?
-         raise NotImplementedError
- 
-diff --git a/gia/eval/rl/scores_dict.json b/gia/eval/rl/scores_dict.json
-index 1b8ebee..ff7d030 100644
---- a/gia/eval/rl/scores_dict.json
-+++ b/gia/eval/rl/scores_dict.json
-@@ -929,8 +929,8 @@
-     },
-     "metaworld-assembly": {
-         "expert": {
--            "mean": 311.29314618777823,
--            "std": 75.04282151450695
-+            "mean": 3523.81468486244,
-+            "std": 63.22745220327798
-         },
-         "random": {
-             "mean": 220.65601680730813,