WIP Add test_backward

tests/test_regression.py

@@ -0,0 +1,81 @@
+import pytest
+import torch
+
+from lerobot.common.datasets.factory import make_dataset
+from lerobot.common.policies.factory import make_policy
+from lerobot.common.utils.utils import init_hydra_config
+from lerobot.scripts.train import make_optimizer
+from tests.utils import DEFAULT_CONFIG_PATH, DEVICE, require_env
+
+
+@pytest.mark.parametrize(
+    "env_name,policy_name,extra_overrides",
+    [
+        # ("xarm", "tdmpc", ["policy.mpc=true"]),
+        # ("pusht", "tdmpc", ["policy.mpc=false"]),
+        ("pusht", "diffusion", []),
+        ("aloha", "act", ["env.task=AlohaInsertion-v0", "dataset.repo_id=lerobot/aloha_sim_insertion_human"]),
+        (
+            "aloha",
+            "act",
+            ["env.task=AlohaInsertion-v0", "dataset.repo_id=lerobot/aloha_sim_insertion_scripted"],
+        ),
+        (
+            "aloha",
+            "act",
+            ["env.task=AlohaTransferCube-v0", "dataset.repo_id=lerobot/aloha_sim_transfer_cube_human"],
+        ),
+        (
+            "aloha",
+            "act",
+            ["env.task=AlohaTransferCube-v0", "dataset.repo_id=lerobot/aloha_sim_transfer_cube_scripted"],
+        ),
+    ],
+)
+@require_env
+def test_backward(env_name, policy_name, extra_overrides):
+    cfg = init_hydra_config(
+        DEFAULT_CONFIG_PATH,
+        overrides=[
+            f"env={env_name}",
+            f"policy={policy_name}",
+            f"device={DEVICE}",
+        ]
+        + extra_overrides,
+    )
+    dataset = make_dataset(cfg)
+    policy = make_policy(cfg, dataset_stats=dataset.stats)
+    policy.train()
+    policy.to(DEVICE)
+    optimizer, lr_scheduler = make_optimizer(cfg, policy)
+
+    dataloader = torch.utils.data.DataLoader(
+        dataset,
+        num_workers=4,
+        batch_size=cfg.policy.batch_size,
+        shuffle=True,
+        pin_memory=torch.device("cpu") != DEVICE,
+        drop_last=True,
+    )
+
+    step = 0
+    done = False
+    training_steps = 1
+    while not done:
+        for batch in dataloader:
+            batch = {k: v.to(DEVICE, non_blocking=True) for k, v in batch.items()}
+            output_dict = policy.forward(batch)
+            loss = output_dict["loss"]
+            loss.backward()
+            optimizer.step()
+            optimizer.zero_grad()
+            step += 1
+            if step >= training_steps:
+                done = True
+                break
+
+
+if __name__ == "__main__":
+    test_backward(
+        "aloha", "act", ["env.task=AlohaInsertion-v0", "dataset.repo_id=lerobot/aloha_sim_insertion_scripted"]
+    )