tangger/lerobot
1
0
Fork 1
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
Thomas Wolf
2024-06-19 10:07:41 +02:00
parent 1cd7ca71a1
commit 33166e1d43
2 changed files with 21 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
lerobot/common/policies/normalize.py
View File

@@ -21,7 +21,6 @@ def create_stats_buffers(
shapes: dict[str, list[int]],
modes: dict[str, str],
stats: dict[str, dict[str, Tensor]] | None = None,
std_epsilon: float = 1e-5,
) -> dict[str, dict[str, nn.ParameterDict]]:
"""
Create buffers per modality (e.g. "observation.image", "action") containing their mean, std, min, max
@@ -79,14 +78,10 @@ def create_stats_buffers(
# https://github.com/huggingface/safetensors/blob/079781fd0dc455ba0fe851e2b4507c33d0c0d407/bindings/python/py_src/safetensors/torch.py#L97.
if mode == "mean_std":
buffer["mean"].data = stats[key]["mean"].clone()
buffer["std"].data = stats[key]["std"].clone().clamp_min(std_epsilon)
buffer["std"].data = stats[key]["std"].clone()
elif mode == "min_max":
buffer["min"].data = stats[key]["min"].clone()
buffer["max"].data = stats[key]["max"].clone()
epsilon = (std_epsilon - (stats[key]["max"] - stats[key]["min"]).abs()).clamp_min(
0
) # To add to have at least std_epsilon between min and max
buffer["max"].data += epsilon
stats_buffers[key] = buffer
return stats_buffers
@@ -134,7 +129,8 @@ class Normalize(nn.Module):
self.shapes = shapes
self.modes = modes
self.stats = stats
stats_buffers = create_stats_buffers(shapes, modes, stats, std_epsilon=std_epsilon)
self.std_epsilon = std_epsilon
stats_buffers = create_stats_buffers(shapes, modes, stats)
for key, buffer in stats_buffers.items():
setattr(self, "buffer_" + key.replace(".", "_"), buffer)
@@ -150,12 +146,15 @@ class Normalize(nn.Module):
std = buffer["std"]
assert not torch.isinf(mean).any(), _no_stats_error_str("mean")
assert not torch.isinf(std).any(), _no_stats_error_str("std")
output_batch[key] = (batch[key] - mean) / std
output_batch[key] = (batch[key] - mean) / std.clamp_min(self.std_epsilon)
elif mode == "min_max":
min = buffer["min"]
max = buffer["max"]
assert not torch.isinf(min).any(), _no_stats_error_str("min")
assert not torch.isinf(max).any(), _no_stats_error_str("max")
# To add to have at least std_epsilon between min and max
epsilon = (self.std_epsilon - (max - min).abs()).clamp_min(0)
max = max + epsilon
# normalize to [0,1]
output_batch[key] = (batch[key] - min) / (max - min)
# normalize to [-1, 1]
@@ -207,8 +206,9 @@ class Unnormalize(nn.Module):
self.shapes = shapes
self.modes = modes
self.stats = stats
self.std_epsilon = std_epsilon
# `self.buffer_observation_state["mean"]` contains `torch.tensor(state_dim)`
stats_buffers = create_stats_buffers(shapes, modes, stats, std_epsilon=std_epsilon)
stats_buffers = create_stats_buffers(shapes, modes, stats)
for key, buffer in stats_buffers.items():
setattr(self, "buffer_" + key.replace(".", "_"), buffer)
@@ -224,12 +224,15 @@ class Unnormalize(nn.Module):
std = buffer["std"]
assert not torch.isinf(mean).any(), _no_stats_error_str("mean")
assert not torch.isinf(std).any(), _no_stats_error_str("std")
output_batch[key] = batch[key] * std + mean
output_batch[key] = batch[key] * std.clamp_min(self.std_epsilon) + mean
elif mode == "min_max":
min = buffer["min"]
max = buffer["max"]
assert not torch.isinf(min).any(), _no_stats_error_str("min")
assert not torch.isinf(max).any(), _no_stats_error_str("max")
# To add to have at least std_epsilon between min and max
epsilon = (self.std_epsilon - (max - min).abs()).clamp_min(0)
max = max + epsilon
output_batch[key] = (batch[key] + 1) / 2
output_batch[key] = output_batch[key] * (max - min) + min
else:

36
tests/test_policies.py
View File

@@ -331,24 +331,14 @@ def test_normalize(insert_temporal_dim):
).all()
assert torch.isclose(
normalize.buffer_action_test_std_cap.std[0],
dataset_stats["action_test_std_cap"]["std"][0],
normalize.buffer_action_test_std_cap.std,
dataset_stats["action_test_std_cap"]["std"],
rtol=0.1,
atol=1e-7,
).all()
assert torch.isclose(
normalize.buffer_action_test_std_cap.std[1], torch.ones(1) * std_epsilon, rtol=0.1, atol=1e-7
).all()
assert torch.isclose(
normalize.buffer_action_test_min_max_cap.max[0] - normalize.buffer_action_test_min_max_cap.min[0],
dataset_stats["action_test_min_max_cap"]["max"][0]
- dataset_stats["action_test_min_max_cap"]["min"][0],
rtol=0.1,
atol=1e-7,
).all()
assert torch.isclose(
normalize.buffer_action_test_min_max_cap.max[1] - normalize.buffer_action_test_min_max_cap.min[1],
torch.ones(1) * std_epsilon,
normalize.buffer_action_test_min_max_cap.max - normalize.buffer_action_test_min_max_cap.min,
dataset_stats["action_test_min_max_cap"]["max"] - dataset_stats["action_test_min_max_cap"]["min"],
rtol=0.1,
atol=1e-7,
).all()
@@ -496,24 +486,14 @@ def test_normalize(insert_temporal_dim):
).all()
assert torch.isclose(
unnormalize.buffer_action_test_std_cap.std[0],
dataset_stats["action_test_std_cap"]["std"][0],
unnormalize.buffer_action_test_std_cap.std,
dataset_stats["action_test_std_cap"]["std"],
rtol=0.1,
atol=1e-7,
).all()
assert torch.isclose(
unnormalize.buffer_action_test_std_cap.std[1], torch.ones(1) * std_epsilon, rtol=0.1, atol=1e-7
).all()
assert torch.isclose(
unnormalize.buffer_action_test_min_max_cap.max[0] - unnormalize.buffer_action_test_min_max_cap.min[0],
dataset_stats["action_test_min_max_cap"]["max"][0]
- dataset_stats["action_test_min_max_cap"]["min"][0],
rtol=0.1,
atol=1e-7,
).all()
assert torch.isclose(
unnormalize.buffer_action_test_min_max_cap.max[1] - unnormalize.buffer_action_test_min_max_cap.min[1],
torch.ones(1) * std_epsilon,
unnormalize.buffer_action_test_min_max_cap.max - unnormalize.buffer_action_test_min_max_cap.min,
dataset_stats["action_test_min_max_cap"]["max"] - dataset_stats["action_test_min_max_cap"]["min"],
rtol=0.1,
atol=1e-7,
).all()