test/test_modules.py

# Copyright (c) Meta Platforms, Inc. and affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
import argparse
import os
import re

from numbers import Number

import numpy as np
import pytest
import torch

if os.getenv("PYTORCH_TEST_FBCODE"):
    from pytorch.rl.test._utils_internal import get_default_devices, retry
    from pytorch.rl.test.mocking_classes import MockBatchedUnLockedEnv
else:
    from _utils_internal import get_default_devices, retry
    from mocking_classes import MockBatchedUnLockedEnv
from packaging import version
from tensordict import TensorDict
from torch import nn
from torchrl.data.tensor_specs import Bounded, Composite
from torchrl.modules import (
    CEMPlanner,
    DTActor,
    GRU,
    GRUCell,
    LSTM,
    LSTMCell,
    MultiAgentConvNet,
    MultiAgentMLP,
    OnlineDTActor,
    QMixer,
    SafeModule,
    TanhModule,
    ValueOperator,
    VDNMixer,
)
from torchrl.modules.distributions.utils import safeatanh, safetanh
from torchrl.modules.models import (
    BatchRenorm1d,
    Conv3dNet,
    ConvNet,
    MLP,
    NoisyLazyLinear,
    NoisyLinear,
)
from torchrl.modules.models.decision_transformer import (
    _has_transformers,
    DecisionTransformer,
)
from torchrl.modules.models.model_based import (
    DreamerActor,
    ObsDecoder,
    ObsEncoder,
    RSSMPosterior,
    RSSMPrior,
    RSSMRollout,
)
from torchrl.modules.models.utils import SquashDims
from torchrl.modules.planners.mppi import MPPIPlanner
from torchrl.objectives.value import TDLambdaEstimator


@pytest.fixture
def double_prec_fixture():
    dtype = torch.get_default_dtype()
    torch.set_default_dtype(torch.double)
    yield
    torch.set_default_dtype(dtype)


class TestMLP:
    @pytest.mark.parametrize("in_features", [3, 10, None])
    @pytest.mark.parametrize("out_features", [3, (3, 10)])
    @pytest.mark.parametrize("depth, num_cells", [(3, 32), (None, (32, 32, 32))])
    @pytest.mark.parametrize(
        "activation_class, activation_kwargs",
        [(nn.ReLU, {"inplace": True}), (nn.ReLU, {}), (nn.PReLU, {})],
    )
    @pytest.mark.parametrize(
        "norm_class, norm_kwargs",
        [
            (nn.LazyBatchNorm1d, {}),
            (nn.BatchNorm1d, {"num_features": 32}),
            (nn.LayerNorm, {"normalized_shape": 32}),
        ],
    )
    @pytest.mark.parametrize("dropout", [0.0, 0.5])
    @pytest.mark.parametrize("bias_last_layer", [True, False])
    @pytest.mark.parametrize("single_bias_last_layer", [True, False])
    @pytest.mark.parametrize("layer_class", [nn.Linear, NoisyLinear])
    @pytest.mark.parametrize("device", get_default_devices())
    def test_mlp(
        self,
        in_features,
        out_features,
        depth,
        num_cells,
        activation_class,
        activation_kwargs,
        dropout,
        bias_last_layer,
        norm_class,
        norm_kwargs,
        single_bias_last_layer,
        layer_class,
        device,
        seed=0,
    ):
        torch.manual_seed(seed)
        batch = 2
        mlp = MLP(
            in_features=in_features,
            out_features=out_features,
            depth=depth,
            num_cells=num_cells,
            activation_class=activation_class,
            activation_kwargs=activation_kwargs,
            norm_class=norm_class,
            norm_kwargs=norm_kwargs,
            dropout=dropout,
            bias_last_layer=bias_last_layer,
            single_bias_last_layer=False,
            layer_class=layer_class,
            device=device,
        )
        if in_features is None:
            in_features = 5
        x = torch.randn(batch, in_features, device=device)
        y = mlp(x)
        out_features = (
            [out_features] if isinstance(out_features, Number) else out_features
        )
        assert y.shape == torch.Size([batch, *out_features])

    def test_kwargs(self):
        def make_activation(shift):
            return lambda x: x + shift

        def layer(*args, **kwargs):
            linear = nn.Linear(*args, **kwargs)
            linear.weight.data.copy_(torch.eye(4))
            return linear

        in_features = 4
        out_features = 4
        num_cells = [4, 4, 4]
        mlp = MLP(
            in_features=in_features,
            out_features=out_features,
            num_cells=num_cells,
            activation_class=make_activation,
            activation_kwargs=[{"shift": 0}, {"shift": 1}, {"shift": 2}],
            layer_class=layer,
            layer_kwargs=[{"bias": False}] * 4,
            bias_last_layer=False,
        )
        x = torch.zeros(4)
        y = mlp(x)
        for i, module in enumerate(mlp.modules()):
            if isinstance(module, nn.Linear):
                assert (module.weight == torch.eye(4)).all(), i
                assert module.bias is None, i
        assert (y == 3).all()


@pytest.mark.parametrize("in_features", [3, 10, None])
@pytest.mark.parametrize(
    "input_size, depth, num_cells, kernel_sizes, strides, paddings, expected_features",
    [(100, None, None, 3, 1, 0, 32 * 94 * 94), (100, 3, 32, 3, 1, 1, 32 * 100 * 100)],
)
@pytest.mark.parametrize(
    "activation_class, activation_kwargs",
    [(nn.ReLU, {"inplace": True}), (nn.ReLU, {}), (nn.PReLU, {})],
)
@pytest.mark.parametrize(
    "norm_class, norm_kwargs",
    [(None, None), (nn.LazyBatchNorm2d, {}), (nn.BatchNorm2d, {"num_features": 32})],
)
@pytest.mark.parametrize("bias_last_layer", [True, False])
@pytest.mark.parametrize(
    "aggregator_class, aggregator_kwargs",
    [(SquashDims, {})],
)
@pytest.mark.parametrize("squeeze_output", [False])
@pytest.mark.parametrize("device", get_default_devices())
@pytest.mark.parametrize("batch", [(2,), (2, 2)])
def test_convnet(
    batch,
    in_features,
    depth,
    num_cells,
    kernel_sizes,
    strides,
    paddings,
    activation_class,
    activation_kwargs,
    norm_class,
    norm_kwargs,
    bias_last_layer,
    aggregator_class,
    aggregator_kwargs,
    squeeze_output,
    device,
    input_size,
    expected_features,
    seed=0,
):
    torch.manual_seed(seed)
    convnet = ConvNet(
        in_features=in_features,
        depth=depth,
        num_cells=num_cells,
        kernel_sizes=kernel_sizes,
        strides=strides,
        paddings=paddings,
        activation_class=activation_class,
        activation_kwargs=activation_kwargs,
        norm_class=norm_class,
        norm_kwargs=norm_kwargs,
        bias_last_layer=bias_last_layer,
        aggregator_class=aggregator_class,
        aggregator_kwargs=aggregator_kwargs,
        squeeze_output=squeeze_output,
        device=device,
    )
    if in_features is None:
        in_features = 5
    x = torch.randn(*batch, in_features, input_size, input_size, device=device)
    y = convnet(x)
    assert y.shape == torch.Size([*batch, expected_features])


class TestConv3d:
    @pytest.mark.parametrize("in_features", [3, 10, None])
    @pytest.mark.parametrize(
        "input_size, depth, num_cells, kernel_sizes, strides, paddings, expected_features",
        [
            (10, None, None, 3, 1, 0, 32 * 4 * 4 * 4),
            (10, 3, 32, 3, 1, 1, 32 * 10 * 10 * 10),
        ],
    )
    @pytest.mark.parametrize(
        "activation_class, activation_kwargs",
        [(nn.ReLU, {"inplace": True}), (nn.ReLU, {}), (nn.PReLU, {})],
    )
    @pytest.mark.parametrize(
        "norm_class, norm_kwargs",
        [
            (None, None),
            (nn.LazyBatchNorm3d, {}),
            (nn.BatchNorm3d, {"num_features": 32}),
        ],
    )
    @pytest.mark.parametrize("bias_last_layer", [True, False])
    @pytest.mark.parametrize(
        "aggregator_class, aggregator_kwargs",
        [(SquashDims, None)],
    )
    @pytest.mark.parametrize("squeeze_output", [False])
    @pytest.mark.parametrize("device", get_default_devices())
    @pytest.mark.parametrize("batch", [(2,), (2, 2)])
    def test_conv3dnet(
        self,
        batch,
        in_features,
        depth,
        num_cells,
        kernel_sizes,
        strides,
        paddings,
        activation_class,
        activation_kwargs,
        norm_class,
        norm_kwargs,
        bias_last_layer,
        aggregator_class,
        aggregator_kwargs,
        squeeze_output,
        device,
        input_size,
        expected_features,
        seed=0,
    ):
        torch.manual_seed(seed)
        conv3dnet = Conv3dNet(
            in_features=in_features,
            depth=depth,
            num_cells=num_cells,
            kernel_sizes=kernel_sizes,
            strides=strides,
            paddings=paddings,
            activation_class=activation_class,
            activation_kwargs=activation_kwargs,
            norm_class=norm_class,
            norm_kwargs=norm_kwargs,
            bias_last_layer=bias_last_layer,
            aggregator_class=aggregator_class,
            aggregator_kwargs=aggregator_kwargs,
            squeeze_output=squeeze_output,
            device=device,
        )
        if in_features is None:
            in_features = 5
        x = torch.randn(
            *batch, in_features, input_size, input_size, input_size, device=device
        )
        y = conv3dnet(x)
        assert y.shape == torch.Size([*batch, expected_features])
        with pytest.raises(ValueError, match="must have at least 4 dimensions"):
            conv3dnet(torch.randn(3, 16, 16))

    def test_errors(self):
        with pytest.raises(
            ValueError, match="Null depth is not permitted with Conv3dNet"
        ):
            conv3dnet = Conv3dNet(
                in_features=5,
                num_cells=32,
                depth=0,
            )
        with pytest.raises(
            ValueError, match="depth=None requires one of the input args"
        ):
            conv3dnet = Conv3dNet(
                in_features=5,
                num_cells=32,
                depth=None,
            )
        with pytest.raises(
            ValueError, match="consider matching or specifying a constant num_cells"
        ):
            conv3dnet = Conv3dNet(
                in_features=5,
                num_cells=[32],
                depth=None,
                kernel_sizes=[3, 3],
            )


@pytest.mark.parametrize(
    "layer_class",
    [
        NoisyLinear,
        NoisyLazyLinear,
    ],
)
@pytest.mark.parametrize("device", get_default_devices())
def test_noisy(layer_class, device, seed=0):
    torch.manual_seed(seed)
    layer = layer_class(3, 4, device=device)
    x = torch.randn(10, 3, device=device)
    y1 = layer(x)
    layer.reset_noise()
    y2 = layer(x)
    y3 = layer(x)
    torch.testing.assert_close(y2, y3)
    with pytest.raises(AssertionError):
        torch.testing.assert_close(y1, y2)


@pytest.mark.parametrize("device", get_default_devices())
@pytest.mark.parametrize("batch_size", [3, 5])
class TestPlanner:
    def test_CEM_model_free_env(self, device, batch_size, seed=1):
        env = MockBatchedUnLockedEnv(device=device)
        torch.manual_seed(seed)
        planner = CEMPlanner(
            env,
            planning_horizon=10,
            optim_steps=2,
            num_candidates=100,
            top_k=2,
        )
        td = env.reset(TensorDict(batch_size=batch_size).to(device))
        td_copy = td.clone()
        td = planner(td)
        assert (
            td.get("action").shape[-len(env.action_spec.shape) :]
            == env.action_spec.shape
        )
        assert env.action_spec.is_in(td.get("action"))

        for key in td.keys():
            if key != "action":
                assert torch.allclose(td[key], td_copy[key])

    def test_MPPI(self, device, batch_size, seed=1):
        torch.manual_seed(seed)
        env = MockBatchedUnLockedEnv(device=device)
        value_net = nn.LazyLinear(1, device=device)
        value_net = ValueOperator(value_net, in_keys=["observation"])
        advantage_module = TDLambdaEstimator(
            gamma=0.99,
            lmbda=0.95,
            value_network=value_net,
        )
        value_net(env.reset())
        planner = MPPIPlanner(
            env,
            advantage_module,
            temperature=1.0,
            planning_horizon=10,
            optim_steps=2,
            num_candidates=100,
            top_k=2,
        )
        td = env.reset(TensorDict(batch_size=batch_size).to(device))
        td_copy = td.clone()
        td = planner(td)
        assert (
            td.get("action").shape[-len(env.action_spec.shape) :]
            == env.action_spec.shape
        )
        assert env.action_spec.is_in(td.get("action"))

        for key in td.keys():
            if key != "action":
                assert torch.allclose(td[key], td_copy[key])


@pytest.mark.parametrize("device", get_default_devices())
@pytest.mark.parametrize("batch_size", [[], [3], [5]])
@pytest.mark.skipif(
    version.parse(torch.__version__) < version.parse("1.11.0"),
    reason="""Dreamer works with batches of null to 2 dimensions. Torch < 1.11
requires one-dimensional batches (for RNN and Conv nets for instance). If you'd like
to see torch < 1.11 supported for dreamer, please submit an issue.""",
)
class TestDreamerComponents:
    @pytest.mark.parametrize("out_features", [3, 5])
    @pytest.mark.parametrize("temporal_size", [[], [2], [4]])
    def test_dreamer_actor(self, device, batch_size, temporal_size, out_features):
        actor = DreamerActor(
            out_features,
        ).to(device)
        emb = torch.randn(*batch_size, *temporal_size, 15, device=device)
        state = torch.randn(*batch_size, *temporal_size, 2, device=device)
        loc, scale = actor(emb, state)
        assert loc.shape == (*batch_size, *temporal_size, out_features)
        assert scale.shape == (*batch_size, *temporal_size, out_features)
        assert torch.all(scale > 0)

    @pytest.mark.parametrize("depth", [32, 64])
    @pytest.mark.parametrize("temporal_size", [[], [2], [4]])
    def test_dreamer_encoder(self, device, temporal_size, batch_size, depth):
        encoder = ObsEncoder(channels=depth).to(device)
        obs = torch.randn(*batch_size, *temporal_size, 3, 64, 64, device=device)
        emb = encoder(obs)
        assert emb.shape == (*batch_size, *temporal_size, depth * 8 * 4)

    @pytest.mark.parametrize("depth", [32, 64])
    @pytest.mark.parametrize("stoch_size", [10, 20])
    @pytest.mark.parametrize("deter_size", [20, 30])
    @pytest.mark.parametrize("temporal_size", [[], [2], [4]])
    def test_dreamer_decoder(
        self, device, batch_size, temporal_size, depth, stoch_size, deter_size
    ):
        decoder = ObsDecoder(channels=depth).to(device)
        stoch_state = torch.randn(
            *batch_size, *temporal_size, stoch_size, device=device
        )
        det_state = torch.randn(*batch_size, *temporal_size, deter_size, device=device)
        obs = decoder(stoch_state, det_state)
        assert obs.shape == (*batch_size, *temporal_size, 3, 64, 64)

    @pytest.mark.parametrize("stoch_size", [10, 20])
    @pytest.mark.parametrize("deter_size", [20, 30])
    @pytest.mark.parametrize("action_size", [3, 6])
    def test_rssm_prior(self, device, batch_size, stoch_size, deter_size, action_size):
        action_spec = Bounded(shape=(action_size,), dtype=torch.float32, low=-1, high=1)
        rssm_prior = RSSMPrior(
            action_spec,
            hidden_dim=stoch_size,
            rnn_hidden_dim=stoch_size,
            state_dim=deter_size,
        ).to(device)
        state = torch.randn(*batch_size, deter_size, device=device)
        action = torch.randn(*batch_size, action_size, device=device)
        belief = torch.randn(*batch_size, stoch_size, device=device)
        prior_mean, prior_std, next_state, belief = rssm_prior(state, belief, action)
        assert prior_mean.shape == (*batch_size, deter_size)
        assert prior_std.shape == (*batch_size, deter_size)
        assert next_state.shape == (*batch_size, deter_size)
        assert belief.shape == (*batch_size, stoch_size)
        assert torch.all(prior_std > 0)

    @pytest.mark.parametrize("stoch_size", [10, 20])
    @pytest.mark.parametrize("deter_size", [20, 30])
    def test_rssm_posterior(self, device, batch_size, stoch_size, deter_size):
        rssm_posterior = RSSMPosterior(
            hidden_dim=stoch_size,
            state_dim=deter_size,
        ).to(device)
        belief = torch.randn(*batch_size, stoch_size, device=device)
        obs_emb = torch.randn(*batch_size, 1024, device=device)
        # Init of lazy linears
        _ = rssm_posterior(belief.clone(), obs_emb.clone())

        torch.manual_seed(0)
        posterior_mean, posterior_std, next_state = rssm_posterior(
            belief.clone(), obs_emb.clone()
        )
        assert posterior_mean.shape == (*batch_size, deter_size)
        assert posterior_std.shape == (*batch_size, deter_size)
        assert next_state.shape == (*batch_size, deter_size)
        assert torch.all(posterior_std > 0)

        torch.manual_seed(0)
        posterior_mean_bis, posterior_std_bis, next_state_bis = rssm_posterior(
            belief.clone(), obs_emb.clone()
        )
        assert torch.allclose(posterior_mean, posterior_mean_bis)
        assert torch.allclose(posterior_std, posterior_std_bis)
        assert torch.allclose(next_state, next_state_bis)

    @pytest.mark.parametrize("stoch_size", [10, 20])
    @pytest.mark.parametrize("deter_size", [20, 30])
    @pytest.mark.parametrize("temporal_size", [2, 4])
    @pytest.mark.parametrize("action_size", [3, 6])
    def test_rssm_rollout(
        self, device, batch_size, temporal_size, stoch_size, deter_size, action_size
    ):
        action_spec = Bounded(shape=(action_size,), dtype=torch.float32, low=-1, high=1)
        rssm_prior = RSSMPrior(
            action_spec,
            hidden_dim=stoch_size,
            rnn_hidden_dim=stoch_size,
            state_dim=deter_size,
        ).to(device)
        rssm_posterior = RSSMPosterior(
            hidden_dim=stoch_size,
            state_dim=deter_size,
        ).to(device)

        rssm_rollout = RSSMRollout(
            SafeModule(
                rssm_prior,
                in_keys=["state", "belief", "action"],
                out_keys=[
                    ("next", "prior_mean"),
                    ("next", "prior_std"),
                    "_",
                    ("next", "belief"),
                ],
            ),
            SafeModule(
                rssm_posterior,
                in_keys=[("next", "belief"), ("next", "encoded_latents")],
                out_keys=[
                    ("next", "posterior_mean"),
                    ("next", "posterior_std"),
                    ("next", "state"),
                ],
            ),
        )

        state = torch.randn(*batch_size, temporal_size, deter_size, device=device)
        belief = torch.randn(*batch_size, temporal_size, stoch_size, device=device)
        action = torch.randn(*batch_size, temporal_size, action_size, device=device)
        obs_emb = torch.randn(*batch_size, temporal_size, 1024, device=device)

        tensordict = TensorDict(
            {
                "state": state.clone(),
                "action": action.clone(),
                "next": {
                    "encoded_latents": obs_emb.clone(),
                    "belief": belief.clone(),
                },
            },
            device=device,
            batch_size=torch.Size([*batch_size, temporal_size]),
        )
        ## Init of lazy linears
        _ = rssm_rollout(tensordict.clone())
        torch.manual_seed(0)
        rollout = rssm_rollout(tensordict)
        assert rollout["next", "prior_mean"].shape == (
            *batch_size,
            temporal_size,
            deter_size,
        )
        assert rollout["next", "prior_std"].shape == (
            *batch_size,
            temporal_size,
            deter_size,
        )
        assert rollout["next", "state"].shape == (
            *batch_size,
            temporal_size,
            deter_size,
        )
        assert rollout["next", "belief"].shape == (
            *batch_size,
            temporal_size,
            stoch_size,
        )
        assert rollout["next", "posterior_mean"].shape == (
            *batch_size,
            temporal_size,
            deter_size,
        )
        assert rollout["next", "posterior_std"].shape == (
            *batch_size,
            temporal_size,
            deter_size,
        )
        assert torch.all(rollout["next", "prior_std"] > 0)
        assert torch.all(rollout["next", "posterior_std"] > 0)

        state[..., 1:, :] = 0
        belief[..., 1:, :] = 0
        # Only the first state is used for the prior. The rest are recomputed

        tensordict_bis = TensorDict(
            {
                "state": state.clone(),
                "action": action.clone(),
                "next": {"encoded_latents": obs_emb.clone(), "belief": belief.clone()},
            },
            device=device,
            batch_size=torch.Size([*batch_size, temporal_size]),
        )
        torch.manual_seed(0)
        rollout_bis = rssm_rollout(tensordict_bis)

        assert torch.allclose(
            rollout["next", "prior_mean"], rollout_bis["next", "prior_mean"]
        ), (rollout["next", "prior_mean"] - rollout_bis["next", "prior_mean"]).norm()
        assert torch.allclose(
            rollout["next", "prior_std"], rollout_bis["next", "prior_std"]
        )
        assert torch.allclose(rollout["next", "state"], rollout_bis["next", "state"])
        assert torch.allclose(rollout["next", "belief"], rollout_bis["next", "belief"])
        assert torch.allclose(
            rollout["next", "posterior_mean"], rollout_bis["next", "posterior_mean"]
        )
        assert torch.allclose(
            rollout["next", "posterior_std"], rollout_bis["next", "posterior_std"]
        )


class TestTanh:
    def test_errors(self):
        with pytest.raises(
            ValueError, match="in_keys and out_keys should have the same length"
        ):
            TanhModule(in_keys=["a", "b"], out_keys=["a"])
        with pytest.raises(ValueError, match=r"The minimum value \(-2\) provided"):
            spec = Bounded(-1, 1, shape=())
            TanhModule(in_keys=["act"], low=-2, spec=spec)
        with pytest.raises(ValueError, match=r"The maximum value \(-2\) provided to"):
            spec = Bounded(-1, 1, shape=())
            TanhModule(in_keys=["act"], high=-2, spec=spec)
        with pytest.raises(ValueError, match="Got high < low"):
            TanhModule(in_keys=["act"], high=-2, low=-1)

    def test_minmax(self):
        mod = TanhModule(
            in_keys=["act"],
            high=2,
        )
        assert isinstance(mod.act_high, torch.Tensor)
        mod = TanhModule(
            in_keys=["act"],
            low=-2,
        )
        assert isinstance(mod.act_low, torch.Tensor)
        mod = TanhModule(
            in_keys=["act"],
            high=np.ones((1,)),
        )
        assert isinstance(mod.act_high, torch.Tensor)
        mod = TanhModule(
            in_keys=["act"],
            low=-np.ones((1,)),
        )
        assert isinstance(mod.act_low, torch.Tensor)

    @pytest.mark.parametrize("clamp", [True, False])
    def test_boundaries(self, clamp):
        torch.manual_seed(0)
        eps = torch.finfo(torch.float).resolution
        for _ in range(10):
            min, max = (5 * torch.randn(2)).sort()[0]
            mod = TanhModule(in_keys=["act"], low=min, high=max, clamp=clamp)
            assert mod.non_trivial
            td = TensorDict({"act": (2 * torch.rand(100) - 1) * 10}, [])
            mod(td)
            # we should have a good proportion of samples close to the boundaries
            assert torch.isclose(td["act"], max).any()
            assert torch.isclose(td["act"], min).any()
            if not clamp:
                assert (td["act"] <= max + eps).all()
                assert (td["act"] >= min - eps).all()
            else:
                assert (td["act"] < max + eps).all()
                assert (td["act"] > min - eps).all()

    @pytest.mark.parametrize("out_keys", [[("a", "c"), "b"], None])
    @pytest.mark.parametrize("has_spec", [[True, True], [True, False], [False, False]])
    def test_multi_inputs(self, out_keys, has_spec):
        in_keys = [("x", "z"), "y"]
        real_out_keys = out_keys if out_keys is not None else in_keys

        if any(has_spec):
            spec = {}
            if has_spec[0]:
                spec.update({real_out_keys[0]: Bounded(-2.0, 2.0, shape=())})
                low, high = -2.0, 2.0
            if has_spec[1]:
                spec.update({real_out_keys[1]: Bounded(-3.0, 3.0, shape=())})
                low, high = None, None
            spec = Composite(spec)
        else:
            spec = None
            low, high = -2.0, 2.0

        mod = TanhModule(
            in_keys=in_keys,
            out_keys=out_keys,
            low=low,
            high=high,
            spec=spec,
            clamp=False,
        )
        data = TensorDict({in_key: torch.randn(100) * 100 for in_key in in_keys}, [])
        mod(data)
        assert all(out_key in data.keys(True, True) for out_key in real_out_keys)
        eps = torch.finfo(torch.float).resolution

        for out_key in real_out_keys:
            key = out_key if isinstance(out_key, str) else "_".join(out_key)
            low_key = f"{key}_low"
            high_key = f"{key}_high"
            min, max = getattr(mod, low_key), getattr(mod, high_key)
            assert torch.isclose(data[out_key], max).any()
            assert torch.isclose(data[out_key], min).any()
            assert (data[out_key] <= max + eps).all()
            assert (data[out_key] >= min - eps).all()


class TestMultiAgent:
    def _get_mock_input_td(
        self, n_agents, n_agents_inputs, state_shape=(64, 64, 3), T=None, batch=(2,)
    ):
        if T is not None:
            batch = batch + (T,)
        obs = torch.randn(*batch, n_agents, n_agents_inputs)
        state = torch.randn(*batch, *state_shape)

        td = TensorDict(
            {
                "agents": TensorDict(
                    {"observation": obs},
                    [*batch, n_agents],
                ),
                "state": state,
            },
            batch_size=batch,
        )
        return td

    @retry(AssertionError, 5)
    @pytest.mark.parametrize("n_agents", [1, 3])
    @pytest.mark.parametrize("share_params", [True, False])
    @pytest.mark.parametrize("centralized", [True, False])
    @pytest.mark.parametrize("n_agent_inputs", [6, None])
    @pytest.mark.parametrize("batch", [(4,), (4, 3), ()])
    def test_multiagent_mlp(
        self,
        n_agents,
        centralized,
        share_params,
        batch,
        n_agent_inputs,
        n_agent_outputs=2,
    ):
        torch.manual_seed(1)
        mlp = MultiAgentMLP(
            n_agent_inputs=n_agent_inputs,
            n_agent_outputs=n_agent_outputs,
            n_agents=n_agents,
            centralized=centralized,
            share_params=share_params,
            depth=2,
        )
        if n_agent_inputs is None:
            n_agent_inputs = 6
        td = self._get_mock_input_td(n_agents, n_agent_inputs, batch=batch)
        obs = td.get(("agents", "observation"))

        out = mlp(obs)
        assert out.shape == (*batch, n_agents, n_agent_outputs)
        for i in range(n_agents):
            if centralized and share_params:
                assert torch.allclose(out[..., i, :], out[..., 0, :])
            else:
                for j in range(i + 1, n_agents):
                    assert not torch.allclose(out[..., i, :], out[..., j, :])

        obs[..., 0, 0] += 1
        out2 = mlp(obs)
        for i in range(n_agents):
            if centralized:
                # a modification to the input of agent 0 will impact all agents
                assert not torch.allclose(out[..., i, :], out2[..., i, :])
            elif i > 0:
                assert torch.allclose(out[..., i, :], out2[..., i, :])

        obs = (
            torch.randn(*batch, 1, n_agent_inputs)
            .expand(*batch, n_agents, n_agent_inputs)
            .clone()
        )
        out = mlp(obs)
        for i in range(n_agents):
            if share_params:
                # same input same output
                assert torch.allclose(out[..., i, :], out[..., 0, :])
            else:
                for j in range(i + 1, n_agents):
                    # same input different output
                    assert not torch.allclose(out[..., i, :], out[..., j, :])
        pattern = rf"""MultiAgentMLP\(
    MLP\(
      \(0\): Linear\(in_features=\d+, out_features=32, bias=True\)
      \(1\): Tanh\(\)
      \(2\): Linear\(in_features=32, out_features=32, bias=True\)
      \(3\): Tanh\(\)
      \(4\): Linear\(in_features=32, out_features=2, bias=True\)
    \),
    n_agents={n_agents},
    share_params={share_params},
    centralized={centralized},
    agent_dim={-2}\)"""
        assert re.match(pattern, str(mlp), re.DOTALL)

    @retry(AssertionError, 5)
    @pytest.mark.parametrize("n_agents", [1, 3])
    @pytest.mark.parametrize("share_params", [True, False])
    @pytest.mark.parametrize("centralized", [True, False])
    @pytest.mark.parametrize("n_agent_inputs", [6, None])
    @pytest.mark.parametrize("batch", [(4,), (4, 3), ()])
    def test_multiagent_mlp_init(
        self,
        n_agents,
        centralized,
        share_params,
        batch,
        n_agent_inputs,
        n_agent_outputs=2,
    ):
        torch.manual_seed(1)
        mlp = MultiAgentMLP(
            n_agent_inputs=n_agent_inputs,
            n_agent_outputs=n_agent_outputs,
            n_agents=n_agents,
            centralized=centralized,
            share_params=share_params,
            depth=2,
        )
        for m in mlp.modules():
            if isinstance(m, nn.Linear):
                assert not isinstance(m.weight, nn.Parameter)
                assert m.weight.device == torch.device("meta")
                break
        else:
            raise RuntimeError("could not find a Linear module")
        if n_agent_inputs is None:
            n_agent_inputs = 6
        td = self._get_mock_input_td(n_agents, n_agent_inputs, batch=batch)
        obs = td.get(("agents", "observation"))
        mlp(obs)
        snet = mlp.get_stateful_net()
        assert snet is not mlp._empty_net

        def zero_inplace(mod):
            if hasattr(mod, "weight"):
                mod.weight.data *= 0
            if hasattr(mod, "bias"):
                mod.bias.data *= 0

        snet.apply(zero_inplace)
        assert (mlp.params == 0).all()

        def one_outofplace(mod):
            if hasattr(mod, "weight"):
                mod.weight = nn.Parameter(torch.ones_like(mod.weight.data))
            if hasattr(mod, "bias"):
                mod.bias = nn.Parameter(torch.ones_like(mod.bias.data))

        snet.apply(one_outofplace)
        assert (mlp.params == 0).all()
        mlp.from_stateful_net(snet)
        assert (mlp.params == 1).all()

    @retry(AssertionError, 5)
    @pytest.mark.parametrize("n_agents", [3])
    @pytest.mark.parametrize("share_params", [True])
    @pytest.mark.parametrize("centralized", [True])
    @pytest.mark.parametrize("n_agent_inputs", [6])
    @pytest.mark.parametrize("batch", [(4,)])
    @pytest.mark.parametrize("tdparams", [True, False])
    def test_multiagent_mlp_tdparams(
        self,
        n_agents,
        centralized,
        share_params,
        batch,
        n_agent_inputs,
        tdparams,
        n_agent_outputs=2,
    ):
        torch.manual_seed(1)
        mlp = MultiAgentMLP(
            n_agent_inputs=n_agent_inputs,
            n_agent_outputs=n_agent_outputs,
            n_agents=n_agents,
            centralized=centralized,
            share_params=share_params,
            depth=2,
            use_td_params=tdparams,
        )
        if tdparams:
            assert list(mlp._empty_net.parameters()) == []
            assert list(mlp.params.parameters()) == list(mlp.parameters())
        else:
            assert list(mlp._empty_net.parameters()) == list(mlp.parameters())
            assert not hasattr(mlp.params, "parameters")
        if torch.backends.mps.is_available():
            device = torch.device("mps")
        elif torch.cuda.is_available():
            device = torch.device("cuda")
        else:
            return
        mlp = nn.Sequential(mlp)
        mlp_device = mlp.to(device)
        param_set = set(mlp.parameters())
        for p in mlp[0].params.values(True, True):
            assert p in param_set

    def test_multiagent_mlp_lazy(self):
        mlp = MultiAgentMLP(
            n_agent_inputs=None,
            n_agent_outputs=6,
            n_agents=3,
            centralized=True,
            share_params=False,
            depth=2,
        )
        optim = torch.optim.SGD(mlp.parameters(), lr=1e-3)
        for p in mlp.parameters():
            if isinstance(p, torch.nn.parameter.UninitializedParameter):
                break
        else:
            raise AssertionError("No UninitializedParameter found")
        for p in optim.param_groups[0]["params"]:
            if isinstance(p, torch.nn.parameter.UninitializedParameter):
                break
        else:
            raise AssertionError("No UninitializedParameter found")
        for _ in range(2):
            td = self._get_mock_input_td(3, 4, batch=(10,))
            obs = td.get(("agents", "observation"))
            out = mlp(obs)
            assert (
                not mlp.params[0]
                .apply(lambda x, y: torch.isclose(x, y), mlp.params[1])
                .any()
            )
            out.mean().backward()
            optim.step()
        for p in mlp.parameters():
            if isinstance(p, torch.nn.parameter.UninitializedParameter):
                raise AssertionError("UninitializedParameter found")
        for p in optim.param_groups[0]["params"]:
            if isinstance(p, torch.nn.parameter.UninitializedParameter):
                raise AssertionError("UninitializedParameter found")

    @pytest.mark.parametrize("n_agents", [1, 3])
    @pytest.mark.parametrize("share_params", [True, False])
    @pytest.mark.parametrize("centralized", [True, False])
    def test_multiagent_reset_mlp(
        self,
        n_agents,
        centralized,
        share_params,
    ):
        actor_net = MultiAgentMLP(
            n_agent_inputs=4,
            n_agent_outputs=6,
            num_cells=(4, 4),
            n_agents=n_agents,
            centralized=centralized,
            share_params=share_params,
        )
        params_before = actor_net.params.clone()
        actor_net.reset_parameters()
        params_after = actor_net.params
        assert not params_before.apply(
            lambda x, y: torch.isclose(x, y), params_after, batch_size=[]
        ).any()
        if params_after.numel() > 1:
            assert (
                not params_after[0]
                .apply(lambda x, y: torch.isclose(x, y), params_after[1], batch_size=[])
                .any()
            )

    @pytest.mark.parametrize("n_agents", [1, 3])
    @pytest.mark.parametrize("share_params", [True, False])
    @pytest.mark.parametrize("centralized", [True, False])
    @pytest.mark.parametrize("channels", [3, None])
    @pytest.mark.parametrize("batch", [(4,), (4, 3), ()])
    def test_multiagent_cnn(
        self,
        n_agents,
        centralized,
        share_params,
        batch,
        channels,
        x=15,
        y=15,
    ):
        torch.manual_seed(0)
        cnn = MultiAgentConvNet(
            n_agents=n_agents,
            centralized=centralized,
            share_params=share_params,
            in_features=channels,
            kernel_sizes=3,
        )
        if channels is None:
            channels = 3
        td = TensorDict(
            {
                "agents": TensorDict(
                    {"observation": torch.randn(*batch, n_agents, channels, x, y)},
                    [*batch, n_agents],
                )
            },
            batch_size=batch,
        )
        obs = td[("agents", "observation")]
        out = cnn(obs)
        assert out.shape[:-1] == (*batch, n_agents)
        if centralized and share_params:
            torch.testing.assert_close(out, out[..., :1, :].expand_as(out))
        else:
            for i in range(n_agents):
                for j in range(i + 1, n_agents):
                    assert not torch.allclose(out[..., i, :], out[..., j, :])
        obs[..., 0, 0, 0, 0] += 1
        out2 = cnn(obs)
        if centralized:
            # a modification to the input of agent 0 will impact all agents
            assert not torch.isclose(out, out2).all()
        elif n_agents > 1:
            assert not torch.isclose(out[..., 0, :], out2[..., 0, :]).all()
            torch.testing.assert_close(out[..., 1:, :], out2[..., 1:, :])

        obs = torch.randn(*batch, 1, channels, x, y).expand(
            *batch, n_agents, channels, x, y
        )
        out = cnn(obs)
        for i in range(n_agents):
            if share_params:
                # same input same output
                assert torch.allclose(out[..., i, :], out[..., 0, :])
            else:
                for j in range(i + 1, n_agents):
                    # same input different output
                    assert not torch.allclose(out[..., i, :], out[..., j, :])

    def test_multiagent_cnn_lazy(self):
        n_agents = 5
        n_channels = 3
        cnn = MultiAgentConvNet(
            n_agents=n_agents,
            centralized=False,
            share_params=False,
            in_features=None,
            kernel_sizes=3,
        )
        optim = torch.optim.SGD(cnn.parameters(), lr=1e-3)
        for p in cnn.parameters():
            if isinstance(p, torch.nn.parameter.UninitializedParameter):
                break
        else:
            raise AssertionError("No UninitializedParameter found")
        for p in optim.param_groups[0]["params"]:
            if isinstance(p, torch.nn.parameter.UninitializedParameter):
                break
        else:
            raise AssertionError("No UninitializedParameter found")
        for _ in range(2):
            td = TensorDict(
                {
                    "agents": TensorDict(
                        {"observation": torch.randn(4, n_agents, n_channels, 15, 15)},
                        [4, 5],
                    )
                },
                batch_size=[4],
            )
            obs = td[("agents", "observation")]
            out = cnn(obs)
            assert (
                not cnn.params[0]
                .apply(lambda x, y: torch.isclose(x, y), cnn.params[1])
                .any()
            )
            out.mean().backward()
            optim.step()
        for p in cnn.parameters():
            if isinstance(p, torch.nn.parameter.UninitializedParameter):
                raise AssertionError("UninitializedParameter found")
        for p in optim.param_groups[0]["params"]:
            if isinstance(p, torch.nn.parameter.UninitializedParameter):
                raise AssertionError("UninitializedParameter found")

    @pytest.mark.parametrize("n_agents", [1, 3])
    @pytest.mark.parametrize("share_params", [True, False])
    @pytest.mark.parametrize("centralized", [True, False])
    def test_multiagent_reset_cnn(
        self,
        n_agents,
        centralized,
        share_params,
    ):
        actor_net = MultiAgentConvNet(
            in_features=4,
            num_cells=[5, 5],
            n_agents=n_agents,
            centralized=centralized,
            share_params=share_params,
        )
        params_before = actor_net.params.clone()
        actor_net.reset_parameters()
        params_after = actor_net.params
        assert not params_before.apply(
            lambda x, y: torch.isclose(x, y), params_after, batch_size=[]
        ).any()
        if params_after.numel() > 1:
            assert (
                not params_after[0]
                .apply(lambda x, y: torch.isclose(x, y), params_after[1], batch_size=[])
                .any()
            )

    @pytest.mark.parametrize("n_agents", [1, 3])
    @pytest.mark.parametrize("batch", [(10,), (10, 3), ()])
    def test_vdn(self, n_agents, batch):
        torch.manual_seed(0)
        mixer = VDNMixer(n_agents=n_agents, device="cpu")

        td = self._get_mock_input_td(n_agents, batch=batch, n_agents_inputs=1)
        obs = td.get(("agents", "observation"))
        assert obs.shape == (*batch, n_agents, 1)
        out = mixer(obs)
        assert out.shape == (*batch, 1)
        assert torch.equal(obs.sum(-2), out)

    @pytest.mark.parametrize("n_agents", [1, 3])
    @pytest.mark.parametrize("batch", [(10,), (10, 3), ()])
    @pytest.mark.parametrize("state_shape", [(64, 64, 3), (10,)])
    def test_qmix(self, n_agents, batch, state_shape):
        torch.manual_seed(0)
        mixer = QMixer(
            n_agents=n_agents,
            state_shape=state_shape,
            mixing_embed_dim=32,
            device="cpu",
        )

        td = self._get_mock_input_td(
            n_agents, batch=batch, n_agents_inputs=1, state_shape=state_shape
        )
        obs = td.get(("agents", "observation"))
        state = td.get("state")
        assert obs.shape == (*batch, n_agents, 1)
        assert state.shape == (*batch, *state_shape)
        out = mixer(obs, state)
        assert out.shape == (*batch, 1)

    @pytest.mark.parametrize("mixer", ["qmix", "vdn"])
    def test_mixer_malformed_input(
        self, mixer, n_agents=3, batch=(32,), state_shape=(64, 64, 3)
    ):
        td = self._get_mock_input_td(
            n_agents, batch=batch, n_agents_inputs=3, state_shape=state_shape
        )
        if mixer == "qmix":
            mixer = QMixer(
                n_agents=n_agents,
                state_shape=state_shape,
                mixing_embed_dim=32,
                device="cpu",
            )
        else:
            mixer = VDNMixer(n_agents=n_agents, device="cpu")
        obs = td.get(("agents", "observation"))
        state = td.get("state")

        if mixer.needs_state:
            with pytest.raises(
                ValueError,
                match="Mixer that needs state was passed more than 2 inputs",
            ):
                mixer(obs)
        else:
            with pytest.raises(
                ValueError,
                match="Mixer that doesn't need state was passed more than 1 input",
            ):
                mixer(obs, state)

        in_put = [obs, state] if mixer.needs_state else [obs]
        with pytest.raises(
            ValueError,
            match="Mixer network expected chosen_action_value with last 2 dimensions",
        ):
            mixer(*in_put)
        if mixer.needs_state:
            state_diff = state.unsqueeze(-1)
            with pytest.raises(
                ValueError,
                match="Mixer network expected state with ending shape",
            ):
                mixer(obs, state_diff)

        td = self._get_mock_input_td(
            n_agents, batch=batch, n_agents_inputs=1, state_shape=state_shape
        )
        obs = td.get(("agents", "observation"))
        state = td.get("state")
        obs = obs.sum(-2)
        in_put = [obs, state] if mixer.needs_state else [obs]
        with pytest.raises(
            ValueError,
            match="Mixer network expected chosen_action_value with last 2 dimensions",
        ):
            mixer(*in_put)

        obs = td.get(("agents", "observation"))
        state = td.get("state")
        in_put = [obs, state] if mixer.needs_state else [obs]
        mixer(*in_put)


@pytest.mark.skipif(torch.__version__ < "2.0", reason="torch 2.0 is required")
@pytest.mark.parametrize("use_vmap", [False, True])
@pytest.mark.parametrize("scale", range(10))
def test_tanh_atanh(use_vmap, scale):
    if use_vmap:
        try:
            from torch import vmap
        except ImportError:
            try:
                from functorch import vmap
            except ImportError:
                raise pytest.skip("functorch not found")

    torch.manual_seed(0)
    x = (torch.randn(10, dtype=torch.double) * scale).requires_grad_(True)
    if not use_vmap:
        y = safetanh(x, 1e-6)
    else:
        y = vmap(safetanh, (0, None))(x, 1e-6)

    if not use_vmap:
        xp = safeatanh(y, 1e-6)
    else:
        xp = vmap(safeatanh, (0, None))(y, 1e-6)

    xp.sum().backward()
    torch.testing.assert_close(x.grad, torch.ones_like(x))


@pytest.mark.skipif(
    not _has_transformers, reason="transformers needed for TestDecisionTransformer"
)
class TestDecisionTransformer:
    def test_init(self):
        DecisionTransformer(
            3,
            4,
        )
        with pytest.raises(TypeError):
            DecisionTransformer(3, 4, config="some_str")
        DecisionTransformer(
            3,
            4,
            config=DecisionTransformer.DTConfig(
                n_layer=2, n_embd=16, n_positions=16, n_inner=16, n_head=2
            ),
        )

    @pytest.mark.parametrize("batch_dims", [[], [3], [3, 4]])
    def test_exec(self, batch_dims, T=5):
        observations = torch.randn(*batch_dims, T, 3)
        actions = torch.randn(*batch_dims, T, 4)
        r2go = torch.randn(*batch_dims, T, 1)
        model = DecisionTransformer(
            3,
            4,
            config=DecisionTransformer.DTConfig(
                n_layer=2, n_embd=16, n_positions=16, n_inner=16, n_head=2
            ),
        )
        out = model(observations, actions, r2go)
        assert out.shape == torch.Size([*batch_dims, T, 16])

    @pytest.mark.parametrize("batch_dims", [[], [3], [3, 4]])
    def test_dtactor(self, batch_dims, T=5):
        dtactor = DTActor(
            3,
            4,
            transformer_config=DecisionTransformer.DTConfig(
                n_layer=2, n_embd=16, n_positions=16, n_inner=16, n_head=2
            ),
        )
        observations = torch.randn(*batch_dims, T, 3)
        actions = torch.randn(*batch_dims, T, 4)
        r2go = torch.randn(*batch_dims, T, 1)
        out = dtactor(observations, actions, r2go)
        assert out.shape == torch.Size([*batch_dims, T, 4])

    @pytest.mark.parametrize("batch_dims", [[], [3], [3, 4]])
    def test_onlinedtactor(self, batch_dims, T=5):
        dtactor = OnlineDTActor(
            3,
            4,
            transformer_config=DecisionTransformer.DTConfig(
                n_layer=2, n_embd=16, n_positions=16, n_inner=16, n_head=2
            ),
        )
        observations = torch.randn(*batch_dims, T, 3)
        actions = torch.randn(*batch_dims, T, 4)
        r2go = torch.randn(*batch_dims, T, 1)
        mu, sig = dtactor(observations, actions, r2go)
        assert mu.shape == torch.Size([*batch_dims, T, 4])
        assert sig.shape == torch.Size([*batch_dims, T, 4])
        assert (dtactor.log_std_min < sig.log()).all()
        assert (dtactor.log_std_max > sig.log()).all()


@pytest.mark.parametrize("device", get_default_devices())
@pytest.mark.parametrize("bias", [True, False])
def test_python_lstm_cell(device, bias):
    lstm_cell1 = LSTMCell(10, 20, device=device, bias=bias)
    lstm_cell2 = nn.LSTMCell(10, 20, device=device, bias=bias)

    lstm_cell1.load_state_dict(lstm_cell2.state_dict())

    # Make sure parameters match
    for (k1, v1), (k2, v2) in zip(
        lstm_cell1.named_parameters(), lstm_cell2.named_parameters()
    ):
        assert k1 == k2, f"Parameter names do not match: {k1} != {k2}"
        assert (
            v1.shape == v2.shape
        ), f"Parameter shapes do not match: {k1} shape {v1.shape} != {k2} shape {v2.shape}"

    # Run loop
    input = torch.randn(2, 3, 10, device=device)
    h0 = torch.randn(3, 20, device=device)
    c0 = torch.randn(3, 20, device=device)
    with torch.no_grad():
        for i in range(input.size()[0]):
            h1, c1 = lstm_cell1(input[i], (h0, c0))
            h2, c2 = lstm_cell2(input[i], (h0, c0))

            # Make sure the final hidden states have the same shape
            assert h1.shape == h2.shape
            assert c1.shape == c2.shape
            torch.testing.assert_close(h1, h2)
            torch.testing.assert_close(c1, c2)
            h0 = h1
            c0 = c1


@pytest.mark.parametrize("device", get_default_devices())
@pytest.mark.parametrize("bias", [True, False])
def test_python_gru_cell(device, bias):
    gru_cell1 = GRUCell(10, 20, device=device, bias=bias)
    gru_cell2 = nn.GRUCell(10, 20, device=device, bias=bias)

    gru_cell2.load_state_dict(gru_cell1.state_dict())

    # Make sure parameters match
    for (k1, v1), (k2, v2) in zip(
        gru_cell1.named_parameters(), gru_cell2.named_parameters()
    ):
        assert k1 == k2, f"Parameter names do not match: {k1} != {k2}"
        assert (v1 == v2).all()
        assert (
            v1.shape == v2.shape
        ), f"Parameter shapes do not match: {k1} shape {v1.shape} != {k2} shape {v2.shape}"

    # Run loop
    input = torch.randn(2, 3, 10, device=device)
    h0 = torch.zeros(3, 20, device=device)
    with torch.no_grad():
        for i in range(input.size()[0]):
            h1 = gru_cell1(input[i], h0)
            h2 = gru_cell2(input[i], h0)

            # Make sure the final hidden states have the same shape
            assert h1.shape == h2.shape
            torch.testing.assert_close(h1, h2)
            h0 = h1


@pytest.mark.parametrize("device", get_default_devices())
@pytest.mark.parametrize("bias", [True, False])
@pytest.mark.parametrize("batch_first", [True, False])
@pytest.mark.parametrize("dropout", [0.0, 0.5])
@pytest.mark.parametrize("num_layers", [1, 2])
def test_python_lstm(device, bias, dropout, batch_first, num_layers):
    B = 5
    T = 3
    lstm1 = LSTM(
        input_size=10,
        hidden_size=20,
        num_layers=num_layers,
        device=device,
        bias=bias,
        batch_first=batch_first,
    )
    lstm2 = nn.LSTM(
        input_size=10,
        hidden_size=20,
        num_layers=num_layers,
        device=device,
        bias=bias,
        batch_first=batch_first,
    )

    lstm2.load_state_dict(lstm1.state_dict())

    # Make sure parameters match
    for (k1, v1), (k2, v2) in zip(lstm1.named_parameters(), lstm2.named_parameters()):
        assert k1 == k2, f"Parameter names do not match: {k1} != {k2}"
        assert (
            v1.shape == v2.shape
        ), f"Parameter shapes do not match: {k1} shape {v1.shape} != {k2} shape {v2.shape}"

    if batch_first:
        input = torch.randn(B, T, 10, device=device)
    else:
        input = torch.randn(T, B, 10, device=device)

    h0 = torch.randn(num_layers, 5, 20, device=device)
    c0 = torch.randn(num_layers, 5, 20, device=device)

    # Test without hidden states
    with torch.no_grad():
        output1, (h1, c1) = lstm1(input)
        output2, (h2, c2) = lstm2(input)

    assert h1.shape == h2.shape
    assert c1.shape == c2.shape
    assert output1.shape == output2.shape
    if dropout == 0.0:
        torch.testing.assert_close(output1, output2)
        torch.testing.assert_close(h1, h2)
        torch.testing.assert_close(c1, c2)

    # Test with hidden states
    with torch.no_grad():
        output1, (h1, c1) = lstm1(input, (h0, c0))
        output2, (h2, c2) = lstm1(input, (h0, c0))

    assert h1.shape == h2.shape
    assert c1.shape == c2.shape
    assert output1.shape == output2.shape
    if dropout == 0.0:
        torch.testing.assert_close(output1, output2)
        torch.testing.assert_close(h1, h2)
        torch.testing.assert_close(c1, c2)


@pytest.mark.parametrize("device", get_default_devices())
@pytest.mark.parametrize("bias", [True, False])
@pytest.mark.parametrize("batch_first", [True, False])
@pytest.mark.parametrize("dropout", [0.0, 0.5])
@pytest.mark.parametrize("num_layers", [1, 2])
def test_python_gru(device, bias, dropout, batch_first, num_layers):
    B = 5
    T = 3
    gru1 = GRU(
        input_size=10,
        hidden_size=20,
        num_layers=num_layers,
        device=device,
        bias=bias,
        batch_first=batch_first,
    )
    gru2 = nn.GRU(
        input_size=10,
        hidden_size=20,
        num_layers=num_layers,
        device=device,
        bias=bias,
        batch_first=batch_first,
    )
    gru2.load_state_dict(gru1.state_dict())

    # Make sure parameters match
    for (k1, v1), (k2, v2) in zip(gru1.named_parameters(), gru2.named_parameters()):
        assert k1 == k2, f"Parameter names do not match: {k1} != {k2}"
        torch.testing.assert_close(v1, v2)
        assert (
            v1.shape == v2.shape
        ), f"Parameter shapes do not match: {k1} shape {v1.shape} != {k2} shape {v2.shape}"

    if batch_first:
        input = torch.randn(B, T, 10, device=device)
    else:
        input = torch.randn(T, B, 10, device=device)

    h0 = torch.randn(num_layers, 5, 20, device=device)

    # Test without hidden states
    with torch.no_grad():
        output1, h1 = gru1(input)
        output2, h2 = gru2(input)

    assert h1.shape == h2.shape
    assert output1.shape == output2.shape
    if dropout == 0.0:
        torch.testing.assert_close(output1, output2)
        torch.testing.assert_close(h1, h2)

    # Test with hidden states
    with torch.no_grad():
        output1, h1 = gru1(input, h0)
        output2, h2 = gru2(input, h0)

    assert h1.shape == h2.shape
    assert output1.shape == output2.shape
    if dropout == 0.0:
        torch.testing.assert_close(output1, output2)
        torch.testing.assert_close(h1, h2)


class TestBatchRenorm:
    @pytest.mark.parametrize("num_steps", [0, 5])
    @pytest.mark.parametrize("smooth", [False, True])
    def test_batchrenorm(self, num_steps, smooth):
        torch.manual_seed(0)
        bn = torch.nn.BatchNorm1d(5, momentum=0.1, eps=1e-5)
        brn = BatchRenorm1d(
            5,
            momentum=0.1,
            eps=1e-5,
            warmup_steps=num_steps,
            max_d=10000,
            max_r=10000,
            smooth=smooth,
        )
        bn.train()
        brn.train()
        data_train = torch.randn(100, 5).split(25)
        data_test = torch.randn(100, 5)
        for i, d in enumerate(data_train):
            b = bn(d)
            a = brn(d)
            if num_steps > 0 and (
                (i < num_steps and not smooth) or (i == 0 and smooth)
            ):
                torch.testing.assert_close(a, b)
            else:
                assert not torch.isclose(a, b).all(), i

        bn.eval()
        brn.eval()
        torch.testing.assert_close(bn(data_test), brn(data_test))


if __name__ == "__main__":
    args, unknown = argparse.ArgumentParser().parse_known_args()
    pytest.main([__file__, "--capture", "no", "--exitfirst"] + unknown)