wangzixiang
/
lerobot
mirror of https://github.com/huggingface/lerobot.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/main' into user/aliberts/2025_02_25_refactor_robots

This commit is contained in:
Simon Alibert 2025-03-10 18:39:48 +01:00
parent 9bd0788131 84565c7c2e
commit d6ccdc222c
135 changed files with 2177 additions and 514 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
.dockerignore
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Misc
.git
tmp

14
.gitattributes vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
*.memmap filter=lfs diff=lfs merge=lfs -text
*.stl filter=lfs diff=lfs merge=lfs -text
*.safetensors filter=lfs diff=lfs merge=lfs -text

14
.github/ISSUE_TEMPLATE/bug-report.yml vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: "\U0001F41B Bug Report"
description: Submit a bug report to help us improve LeRobot
body:

14
.github/workflows/build-docker-images.yml vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Inspired by
# https://github.com/huggingface/peft/blob/main/.github/workflows/build_docker_images.yml
name: Builds

14
.github/workflows/nightly-tests.yml vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Inspired by
# https://github.com/huggingface/peft/blob/main/.github/workflows/nightly.yml
name: Nightly

125
.github/workflows/pr_style_bot.yml vendored
View File

@ -1,125 +0,0 @@
# Adapted from https://github.com/huggingface/diffusers/blob/main/.github/workflows/pr_style_bot.yml
name: PR Style Bot
on:
issue_comment:
types: [created]
permissions:
contents: write
pull-requests: write
jobs:
run-style-bot:
if: >
contains(github.event.comment.body, '@bot /style') &&
github.event.issue.pull_request != null
runs-on: ubuntu-latest
steps:
- name: Extract PR details
id: pr_info
uses: actions/github-script@v6
with:
script: |
const prNumber = context.payload.issue.number;
const { data: pr } = await github.rest.pulls.get({
owner: context.repo.owner,
repo: context.repo.repo,
pull_number: prNumber
});
// We capture both the branch ref and the "full_name" of the head repo
// so that we can check out the correct repository & branch (including forks).
core.setOutput("prNumber", prNumber);
core.setOutput("headRef", pr.head.ref);
core.setOutput("headRepoFullName", pr.head.repo.full_name);
- name: Check out PR branch
uses: actions/checkout@v4
env:
HEADREPOFULLNAME: ${{ steps.pr_info.outputs.headRepoFullName }}
HEADREF: ${{ steps.pr_info.outputs.headRef }}
with:
persist-credentials: false
# Instead of checking out the base repo, use the contributor's repo name
repository: ${{ env.HEADREPOFULLNAME }}
ref: ${{ env.HEADREF }}
# You may need fetch-depth: 0 for being able to push
fetch-depth: 0
token: ${{ secrets.GITHUB_TOKEN }}
- name: Debug
env:
HEADREPOFULLNAME: ${{ steps.pr_info.outputs.headRepoFullName }}
HEADREF: ${{ steps.pr_info.outputs.headRef }}
PRNUMBER: ${{ steps.pr_info.outputs.prNumber }}
run: |
echo "PR number: $PRNUMBER"
echo "Head Ref: $HEADREF"
echo "Head Repo Full Name: $HEADREPOFULLNAME"
- name: Set up Python
uses: actions/setup-python@v4
- name: Get Ruff Version from pre-commit-config.yaml
id: get-ruff-version
run: |
RUFF_VERSION=$(awk '/repo: https:\/\/github.com\/astral-sh\/ruff-pre-commit/{flag=1;next}/rev:/{if(flag){print $2;exit}}' .pre-commit-config.yaml)
echo "ruff_version=${RUFF_VERSION}" >> $GITHUB_OUTPUT
- name: Install Ruff
env:
RUFF_VERSION: ${{ steps.get-ruff-version.outputs.ruff_version }}
run: python -m pip install "ruff==${RUFF_VERSION}"
- name: Ruff check
run: ruff check --fix
- name: Ruff format
run: ruff format
- name: Commit and push changes
id: commit_and_push
env:
HEADREPOFULLNAME: ${{ steps.pr_info.outputs.headRepoFullName }}
HEADREF: ${{ steps.pr_info.outputs.headRef }}
PRNUMBER: ${{ steps.pr_info.outputs.prNumber }}
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: |
echo "HEADREPOFULLNAME: $HEADREPOFULLNAME, HEADREF: $HEADREF"
# Configure git with the Actions bot user
git config user.name "github-actions[bot]"
git config user.email "github-actions[bot]@users.noreply.github.com"
# Make sure your 'origin' remote is set to the contributor's fork
git remote set-url origin "https://x-access-token:${GITHUB_TOKEN}@github.com/$HEADREPOFULLNAME.git"
# If there are changes after running style/quality, commit them
if [ -n "$(git status --porcelain)" ]; then
git add .
git commit -m "Apply style fixes"
# Push to the original contributor's forked branch
git push origin HEAD:$HEADREF
echo "changes_pushed=true" >> $GITHUB_OUTPUT
else
echo "No changes to commit."
echo "changes_pushed=false" >> $GITHUB_OUTPUT
fi
- name: Comment on PR with workflow run link
if: steps.commit_and_push.outputs.changes_pushed == 'true'
uses: actions/github-script@v6
with:
script: |
const prNumber = parseInt(process.env.prNumber, 10);
const runUrl = `${process.env.GITHUB_SERVER_URL}/${process.env.GITHUB_REPOSITORY}/actions/runs/${process.env.GITHUB_RUN_ID}`
await github.rest.issues.createComment({
owner: context.repo.owner,
repo: context.repo.repo,
issue_number: prNumber,
body: `Style fixes have been applied. [View the workflow run here](${runUrl}).`
});
env:
prNumber: ${{ steps.pr_info.outputs.prNumber }}

14
.github/workflows/quality.yml vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: Quality
on:

14
.github/workflows/test-docker-build.yml vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Inspired by
# https://github.com/huggingface/peft/blob/main/.github/workflows/test-docker-build.yml
name: Test Dockerfiles

14
.github/workflows/test.yml vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: Tests
on:

14
.github/workflows/trufflehog.yml vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
on:
push:

14
.gitignore vendored
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Logging
logs
tmp

31
.pre-commit-config.yaml
View File

@ -1,7 +1,22 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
exclude: ^(tests/data)
default_language_version:
python: python3.10
repos:
##### Style / Misc. #####
- repo: https://github.com/pre-commit/pre-commit-hooks
rev: v5.0.0
hooks:
@ -14,7 +29,7 @@ repos:
- id: end-of-file-fixer
- id: trailing-whitespace
- repo: https://github.com/crate-ci/typos
rev: v1.29.10
rev: v1.30.0
hooks:
- id: typos
args: [--force-exclude]
@ -23,16 +38,24 @@ repos:
hooks:
- id: pyupgrade
- repo: https://github.com/astral-sh/ruff-pre-commit
rev: v0.9.6
rev: v0.9.9
hooks:
- id: ruff
args: [--fix]
- id: ruff-format
##### Security #####
- repo: https://github.com/gitleaks/gitleaks
rev: v8.23.3
rev: v8.24.0
hooks:
- id: gitleaks
- repo: https://github.com/woodruffw/zizmor-pre-commit
rev: v1.3.1
rev: v1.4.1
hooks:
- id: zizmor
- repo: https://github.com/PyCQA/bandit
rev: 1.8.3
hooks:
- id: bandit
args: ["-c", "pyproject.toml"]
additional_dependencies: ["bandit[toml]"]

32
Makefile
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
.PHONY: tests
PYTHON_PATH := $(shell which python)
@ -33,6 +47,7 @@ test-act-ete-train:
--policy.dim_model=64 \
--policy.n_action_steps=20 \
--policy.chunk_size=20 \
--policy.device=$(DEVICE) \
--env.type=aloha \
--env.episode_length=5 \
--dataset.repo_id=lerobot/aloha_sim_transfer_cube_human \
@ -47,7 +62,6 @@ test-act-ete-train:
--save_checkpoint=true \
--log_freq=1 \
--wandb.enable=false \
--device=$(DEVICE) \
--output_dir=tests/outputs/act/
test-act-ete-train-resume:
@ -58,11 +72,11 @@ test-act-ete-train-resume:
test-act-ete-eval:
python lerobot/scripts/eval.py \
--policy.path=tests/outputs/act/checkpoints/000004/pretrained_model \
--policy.device=$(DEVICE) \
--env.type=aloha \
--env.episode_length=5 \
--eval.n_episodes=1 \
--eval.batch_size=1 \
--device=$(DEVICE)
--eval.batch_size=1
test-diffusion-ete-train:
python lerobot/scripts/train.py \
@ -70,6 +84,7 @@ test-diffusion-ete-train:
--policy.down_dims='[64,128,256]' \
--policy.diffusion_step_embed_dim=32 \
--policy.num_inference_steps=10 \
--policy.device=$(DEVICE) \
--env.type=pusht \
--env.episode_length=5 \
--dataset.repo_id=lerobot/pusht \
@ -84,21 +99,21 @@ test-diffusion-ete-train:
--save_freq=2 \
--log_freq=1 \
--wandb.enable=false \
--device=$(DEVICE) \
--output_dir=tests/outputs/diffusion/
test-diffusion-ete-eval:
python lerobot/scripts/eval.py \
--policy.path=tests/outputs/diffusion/checkpoints/000002/pretrained_model \
--policy.device=$(DEVICE) \
--env.type=pusht \
--env.episode_length=5 \
--eval.n_episodes=1 \
--eval.batch_size=1 \
--device=$(DEVICE)
--eval.batch_size=1
test-tdmpc-ete-train:
python lerobot/scripts/train.py \
--policy.type=tdmpc \
--policy.device=$(DEVICE) \
--env.type=xarm \
--env.task=XarmLift-v0 \
--env.episode_length=5 \
@ -114,15 +129,14 @@ test-tdmpc-ete-train:
--save_freq=2 \
--log_freq=1 \
--wandb.enable=false \
--device=$(DEVICE) \
--output_dir=tests/outputs/tdmpc/
test-tdmpc-ete-eval:
python lerobot/scripts/eval.py \
--policy.path=tests/outputs/tdmpc/checkpoints/000002/pretrained_model \
--policy.device=$(DEVICE) \
--env.type=xarm \
--env.episode_length=5 \
--env.task=XarmLift-v0 \
--eval.n_episodes=1 \
--eval.batch_size=1 \
--device=$(DEVICE)
--eval.batch_size=1

22
README.md
View File

@ -23,15 +23,24 @@
</div>
<h2 align="center">
<p><a href="https://github.com/huggingface/lerobot/blob/main/examples/10_use_so100.md">New robot in town: SO-100</a></p>
<p><a href="https://github.com/huggingface/lerobot/blob/main/examples/10_use_so100.md">
Build Your Own SO-100 Robot!</a></p>
</h2>
<div align="center">
<img src="media/so100/leader_follower.webp?raw=true" alt="SO-100 leader and follower arms" title="SO-100 leader and follower arms" width="50%">
<p>We just added a new tutorial on how to build a more affordable robot, at the price of $110 per arm!</p>
<p>Teach it new skills by showing it a few moves with just a laptop.</p>
<p>Then watch your homemade robot act autonomously 🤯</p>
<p>Follow the link to the <a href="https://github.com/huggingface/lerobot/blob/main/examples/10_use_so100.md">full tutorial for SO-100</a>.</p>
<p><strong>Meet the SO-100 – Just $110 per arm!</strong></p>
<p>Train it in minutes with a few simple moves on your laptop.</p>
<p>Then sit back and watch your creation act autonomously! 🤯</p>
<p><a href="https://github.com/huggingface/lerobot/blob/main/examples/10_use_so100.md">
Get the full SO-100 tutorial here.</a></p>
<p>Want to take it to the next level? Make your SO-100 mobile by building LeKiwi!</p>
<p>Check out the <a href="https://github.com/huggingface/lerobot/blob/main/examples/11_use_lekiwi.md">LeKiwi tutorial</a> and bring your robot to life on wheels.</p>
<img src="media/lekiwi/kiwi.webp?raw=true" alt="LeKiwi mobile robot" title="LeKiwi mobile robot" width="50%">
</div>
<br/>
@ -375,3 +384,6 @@ Additionally, if you are using any of the particular policy architecture, pretra
year={2024}
}
```
## Star History
[![Star History Chart](https://api.star-history.com/svg?repos=huggingface/lerobot&type=Timeline)](https://star-history.com/#huggingface/lerobot&Timeline)

14
examples/1_load_lerobot_dataset.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This script demonstrates the use of `LeRobotDataset` class for handling and processing robotic datasets from Hugging Face.
It illustrates how to load datasets, manipulate them, and apply transformations suitable for machine learning tasks in PyTorch.

16
examples/2_evaluate_pretrained_policy.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This scripts demonstrates how to evaluate a pretrained policy from the HuggingFace Hub or from your local
training outputs directory. In the latter case, you might want to run examples/3_train_policy.py first.
@ -30,7 +44,7 @@ pretrained_policy_path = "lerobot/diffusion_pusht"
# OR a path to a local outputs/train folder.
# pretrained_policy_path = Path("outputs/train/example_pusht_diffusion")
policy = DiffusionPolicy.from_pretrained(pretrained_policy_path, map_location=device)
policy = DiffusionPolicy.from_pretrained(pretrained_policy_path)
# Initialize evaluation environment to render two observation types:
# an image of the scene and state/position of the agent. The environment

16
examples/3_train_policy.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""This scripts demonstrates how to train Diffusion Policy on the PushT environment.
Once you have trained a model with this script, you can try to evaluate it on
@ -85,7 +99,7 @@ def main():
done = False
while not done:
for batch in dataloader:
batch = {k: v.to(device, non_blocking=True) for k, v in batch.items()}
batch = {k: (v.to(device) if isinstance(v, torch.Tensor) else v) for k, v in batch.items()}
loss, _ = policy.forward(batch)
loss.backward()
optimizer.step()

14
examples/advanced/1_add_image_transforms.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This script demonstrates how to use torchvision's image transformation with LeRobotDataset for data
augmentation purposes. The transformations are passed to the dataset as an argument upon creation, and

14
examples/advanced/2_calculate_validation_loss.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""This script demonstrates how to slice a dataset and calculate the loss on a subset of the data.
This technique can be useful for debugging and testing purposes, as well as identifying whether a policy

14
examples/port_datasets/pusht_zarr.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import shutil
from pathlib import Path

58
lerobot/common/cameras/intel/camera_realsense.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This file contains utilities for recording frames from Intel Realsense cameras.
"""
@ -99,7 +113,7 @@ def save_images_from_cameras(
camera = RealSenseCamera(config)
camera.connect()
print(
f"RealSenseCamera({camera.serial_number}, fps={camera.fps}, width={camera.width}, height={camera.height}, color_mode={camera.color_mode})"
f"RealSenseCamera({camera.serial_number}, fps={camera.fps}, width={camera.capture_width}, height={camera.capture_height}, color_mode={camera.color_mode})"
)
cameras.append(camera)
@ -209,9 +223,20 @@ class RealSenseCamera(Camera):
self.serial_number = self.find_serial_number_from_name(config.name)
else:
self.serial_number = config.serial_number
self.fps = config.fps
# Store the raw (capture) resolution from the config.
self.capture_width = config.width
self.capture_height = config.height
# If rotated by ±90, swap width and height.
if config.rotation in [-90, 90]:
self.width = config.height
self.height = config.width
else:
self.width = config.width
self.height = config.height
self.fps = config.fps
self.channels = config.channels
self.color_mode = config.color_mode
self.use_depth = config.use_depth
@ -231,7 +256,6 @@ class RealSenseCamera(Camera):
else:
import cv2
# TODO(alibets): Do we keep original width/height or do we define them after rotation?
self.rotation = None
if config.rotation == -90:
self.rotation = cv2.ROTATE_90_COUNTERCLOCKWISE
@ -267,15 +291,19 @@ class RealSenseCamera(Camera):
config = rs.config()
config.enable_device(str(self.serial_number))
if self.fps and self.width and self.height:
if self.fps and self.capture_width and self.capture_height:
# TODO(rcadene): can we set rgb8 directly?
config.enable_stream(rs.stream.color, self.width, self.height, rs.format.rgb8, self.fps)
config.enable_stream(
rs.stream.color, self.capture_width, self.capture_height, rs.format.rgb8, self.fps
)
else:
config.enable_stream(rs.stream.color)
if self.use_depth:
if self.fps and self.width and self.height:
config.enable_stream(rs.stream.depth, self.width, self.height, rs.format.z16, self.fps)
if self.fps and self.capture_width and self.capture_height:
config.enable_stream(
rs.stream.depth, self.capture_width, self.capture_height, rs.format.z16, self.fps
)
else:
config.enable_stream(rs.stream.depth)
@ -313,18 +341,18 @@ class RealSenseCamera(Camera):
raise OSError(
f"Can't set {self.fps=} for RealSenseCamera({self.serial_number}). Actual value is {actual_fps}."
)
if self.width is not None and self.width != actual_width:
if self.capture_width is not None and self.capture_width != actual_width:
raise OSError(
f"Can't set {self.width=} for RealSenseCamera({self.serial_number}). Actual value is {actual_width}."
f"Can't set {self.capture_width=} for RealSenseCamera({self.serial_number}). Actual value is {actual_width}."
)
if self.height is not None and self.height != actual_height:
if self.capture_height is not None and self.capture_height != actual_height:
raise OSError(
f"Can't set {self.height=} for RealSenseCamera({self.serial_number}). Actual value is {actual_height}."
f"Can't set {self.capture_height=} for RealSenseCamera({self.serial_number}). Actual value is {actual_height}."
)
self.fps = round(actual_fps)
self.width = round(actual_width)
self.height = round(actual_height)
self.capture_width = round(actual_width)
self.capture_height = round(actual_height)
self.is_connected = True
@ -370,7 +398,7 @@ class RealSenseCamera(Camera):
color_image = cv2.cvtColor(color_image, cv2.COLOR_RGB2BGR)
h, w, _ = color_image.shape
if h != self.height or w != self.width:
if h != self.capture_height or w != self.capture_width:
raise OSError(
f"Can't capture color image with expected height and width ({self.height} x {self.width}). ({h} x {w}) returned instead."
)
@ -392,7 +420,7 @@ class RealSenseCamera(Camera):
depth_map = np.asanyarray(depth_frame.get_data())
h, w = depth_map.shape
if h != self.height or w != self.width:
if h != self.capture_height or w != self.capture_width:
raise OSError(
f"Can't capture depth map with expected height and width ({self.height} x {self.width}). ({h} x {w}) returned instead."
)

14
lerobot/common/cameras/intel/configuration_realsense.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from dataclasses import dataclass
from ..configs import CameraConfig

72
lerobot/common/cameras/opencv/camera_opencv.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This file contains utilities for recording frames from cameras. For more info look at `OpenCVCamera` docstring.
"""
@ -131,8 +145,8 @@ def save_images_from_cameras(
camera = OpenCVCamera(config)
camera.connect()
print(
f"OpenCVCamera({camera.camera_index}, fps={camera.fps}, width={camera.width}, "
f"height={camera.height}, color_mode={camera.color_mode})"
f"OpenCVCamera({camera.camera_index}, fps={camera.fps}, width={camera.capture_width}, "
f"height={camera.capture_height}, color_mode={camera.color_mode})"
)
cameras.append(camera)
@ -231,9 +245,19 @@ class OpenCVCamera(Camera):
else:
raise ValueError(f"Please check the provided camera_index: {self.camera_index}")
self.fps = config.fps
# Store the raw (capture) resolution from the config.
self.capture_width = config.width
self.capture_height = config.height
# If rotated by ±90, swap width and height.
if config.rotation in [-90, 90]:
self.width = config.height
self.height = config.width
else:
self.width = config.width
self.height = config.height
self.fps = config.fps
self.channels = config.channels
self.color_mode = config.color_mode
self.mock = config.mock
@ -250,7 +274,6 @@ class OpenCVCamera(Camera):
else:
import cv2
# TODO(aliberts): Do we keep original width/height or do we define them after rotation?
self.rotation = None
if config.rotation == -90:
self.rotation = cv2.ROTATE_90_COUNTERCLOCKWISE
@ -272,10 +295,20 @@ class OpenCVCamera(Camera):
# when other threads are used to save the images.
cv2.setNumThreads(1)
backend = (
cv2.CAP_V4L2
if platform.system() == "Linux"
else cv2.CAP_DSHOW
if platform.system() == "Windows"
else cv2.CAP_AVFOUNDATION
if platform.system() == "Darwin"
else cv2.CAP_ANY
)
camera_idx = f"/dev/video{self.camera_index}" if platform.system() == "Linux" else self.camera_index
# First create a temporary camera trying to access `camera_index`,
# and verify it is a valid camera by calling `isOpened`.
tmp_camera = cv2.VideoCapture(camera_idx)
tmp_camera = cv2.VideoCapture(camera_idx, backend)
is_camera_open = tmp_camera.isOpened()
# Release camera to make it accessible for `find_camera_indices`
tmp_camera.release()
@ -298,14 +331,14 @@ class OpenCVCamera(Camera):
# Secondly, create the camera that will be used downstream.
# Note: For some unknown reason, calling `isOpened` blocks the camera which then
# needs to be re-created.
self.camera = cv2.VideoCapture(camera_idx)
self.camera = cv2.VideoCapture(camera_idx, backend)
if self.fps is not None:
self.camera.set(cv2.CAP_PROP_FPS, self.fps)
if self.width is not None:
self.camera.set(cv2.CAP_PROP_FRAME_WIDTH, self.width)
if self.height is not None:
self.camera.set(cv2.CAP_PROP_FRAME_HEIGHT, self.height)
if self.capture_width is not None:
self.camera.set(cv2.CAP_PROP_FRAME_WIDTH, self.capture_width)
if self.capture_height is not None:
self.camera.set(cv2.CAP_PROP_FRAME_HEIGHT, self.capture_height)
actual_fps = self.camera.get(cv2.CAP_PROP_FPS)
actual_width = self.camera.get(cv2.CAP_PROP_FRAME_WIDTH)
@ -317,19 +350,22 @@ class OpenCVCamera(Camera):
raise OSError(
f"Can't set {self.fps=} for OpenCVCamera({self.camera_index}). Actual value is {actual_fps}."
)
if self.width is not None and not math.isclose(self.width, actual_width, rel_tol=1e-3):
if self.capture_width is not None and not math.isclose(
self.capture_width, actual_width, rel_tol=1e-3
):
raise OSError(
f"Can't set {self.width=} for OpenCVCamera({self.camera_index}). Actual value is {actual_width}."
f"Can't set {self.capture_width=} for OpenCVCamera({self.camera_index}). Actual value is {actual_width}."
)
if self.height is not None and not math.isclose(self.height, actual_height, rel_tol=1e-3):
if self.capture_height is not None and not math.isclose(
self.capture_height, actual_height, rel_tol=1e-3
):
raise OSError(
f"Can't set {self.height=} for OpenCVCamera({self.camera_index}). Actual value is {actual_height}."
f"Can't set {self.capture_height=} for OpenCVCamera({self.camera_index}). Actual value is {actual_height}."
)
self.fps = round(actual_fps)
self.width = round(actual_width)
self.height = round(actual_height)
self.capture_width = round(actual_width)
self.capture_height = round(actual_height)
self.is_connected = True
def read(self, temporary_color_mode: str | None = None) -> np.ndarray:
@ -370,7 +406,7 @@ class OpenCVCamera(Camera):
color_image = cv2.cvtColor(color_image, cv2.COLOR_BGR2RGB)
h, w, _ = color_image.shape
if h != self.height or w != self.width:
if h != self.capture_height or w != self.capture_width:
raise OSError(
f"Can't capture color image with expected height and width ({self.height} x {self.width}). ({h} x {w}) returned instead."
)

13
lerobot/common/constants.py
View File

@ -1,3 +1,16 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# keys
import os
from pathlib import Path

14
lerobot/common/datasets/backward_compatibility.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import packaging.version
V2_MESSAGE = """

8
lerobot/common/datasets/lerobot_dataset.py
View File

@ -13,6 +13,7 @@
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import contextlib
import logging
import shutil
from pathlib import Path
@ -27,6 +28,7 @@ import torch.utils
from datasets import concatenate_datasets, load_dataset
from huggingface_hub import HfApi, snapshot_download
from huggingface_hub.constants import REPOCARD_NAME
from huggingface_hub.errors import RevisionNotFoundError
from lerobot.common.constants import HF_LEROBOT_HOME
from lerobot.common.datasets.compute_stats import aggregate_stats, compute_episode_stats
@ -517,6 +519,7 @@ class LeRobotDataset(torch.utils.data.Dataset):
branch: str | None = None,
tags: list | None = None,
license: str | None = "apache-2.0",
tag_version: bool = True,
push_videos: bool = True,
private: bool = False,
allow_patterns: list[str] | str | None = None,
@ -562,6 +565,11 @@ class LeRobotDataset(torch.utils.data.Dataset):
)
card.push_to_hub(repo_id=self.repo_id, repo_type="dataset", revision=branch)
if tag_version:
with contextlib.suppress(RevisionNotFoundError):
hub_api.delete_tag(self.repo_id, tag=CODEBASE_VERSION, repo_type="dataset")
hub_api.create_tag(self.repo_id, tag=CODEBASE_VERSION, revision=branch, repo_type="dataset")
def pull_from_repo(
self,
allow_patterns: list[str] | str | None = None,

14
lerobot/common/datasets/utils.py
View File

@ -31,6 +31,7 @@ import packaging.version
import torch
from datasets.table import embed_table_storage
from huggingface_hub import DatasetCard, DatasetCardData, HfApi
from huggingface_hub.errors import RevisionNotFoundError
from PIL import Image as PILImage
from torchvision import transforms
@ -325,6 +326,19 @@ def get_safe_version(repo_id: str, version: str | packaging.version.Version) ->
)
hub_versions = get_repo_versions(repo_id)
if not hub_versions:
raise RevisionNotFoundError(
f"""Your dataset must be tagged with a codebase version.
Assuming _version_ is the codebase_version value in the info.json, you can run this:
```python
from huggingface_hub import HfApi
hub_api = HfApi()
hub_api.create_tag("{repo_id}", tag="_version_", repo_type="dataset")
```
"""
)
if target_version in hub_versions:
return f"v{target_version}"

14
lerobot/common/datasets/v21/_remove_language_instruction.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import logging
import traceback
from pathlib import Path

16
lerobot/common/datasets/v21/convert_dataset_v20_to_v21.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This script will help you convert any LeRobot dataset already pushed to the hub from codebase version 2.0 to
2.1. It will:
@ -57,7 +71,7 @@ def convert_dataset(
dataset.meta.info["codebase_version"] = CODEBASE_VERSION
write_info(dataset.meta.info, dataset.root)
dataset.push_to_hub(branch=branch, allow_patterns="meta/")
dataset.push_to_hub(branch=branch, tag_version=False, allow_patterns="meta/")
# delete old stats.json file
if (dataset.root / STATS_PATH).is_file:

14
lerobot/common/datasets/v21/convert_stats.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from concurrent.futures import ThreadPoolExecutor, as_completed
import numpy as np

14
lerobot/common/envs/__init__.py
View File

@ -1 +1,15 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from .configs import AlohaEnv, EnvConfig, PushtEnv, XarmEnv # noqa: F401

14
lerobot/common/envs/configs.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import abc
from dataclasses import dataclass, field

14
lerobot/common/motors/configs.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import abc
from dataclasses import dataclass

14
lerobot/common/motors/dynamixel/dynamixel.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import enum
import logging
import math

14
lerobot/common/motors/dynamixel/dynamixel_calibration.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Logic to calibrate a robot arm built with dynamixel motors"""
# TODO(rcadene, aliberts): move this logic into the robot code when refactoring

14
lerobot/common/motors/feetech/feetech.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import enum
import logging
import math

14
lerobot/common/motors/feetech/feetech_calibration.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Logic to calibrate a robot arm built with feetech motors"""
# TODO(rcadene, aliberts): move this logic into the robot code when refactoring

14
lerobot/common/motors/utils.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from .configs import MotorsBusConfig
from .motors_bus import MotorsBus

14
lerobot/common/optim/__init__.py
View File

@ -1 +1,15 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from .optimizers import OptimizerConfig as OptimizerConfig

14
lerobot/common/policies/__init__.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from .act.configuration_act import ACTConfig as ACTConfig
from .diffusion.configuration_diffusion import DiffusionConfig as DiffusionConfig
from .pi0.configuration_pi0 import PI0Config as PI0Config

9
lerobot/common/policies/factory.py
View File

@ -16,7 +16,6 @@
import logging
import torch
from torch import nn
from lerobot.common.datasets.lerobot_dataset import LeRobotDatasetMetadata
@ -76,7 +75,6 @@ def make_policy_config(policy_type: str, **kwargs) -> PreTrainedConfig:
def make_policy(
cfg: PreTrainedConfig,
device: str | torch.device,
ds_meta: LeRobotDatasetMetadata | None = None,
env_cfg: EnvConfig | None = None,
) -> PreTrainedPolicy:
@ -88,7 +86,6 @@ def make_policy(
Args:
cfg (PreTrainedConfig): The config of the policy to make. If `pretrained_path` is set, the policy will
be loaded with the weights from that path.
device (str): the device to load the policy onto.
ds_meta (LeRobotDatasetMetadata | None, optional): Dataset metadata to take input/output shapes and
statistics to use for (un)normalization of inputs/outputs in the policy. Defaults to None.
env_cfg (EnvConfig | None, optional): The config of a gym environment to parse features from. Must be
@ -96,7 +93,7 @@ def make_policy(
Raises:
ValueError: Either ds_meta or env and env_cfg must be provided.
NotImplementedError: if the policy.type is 'vqbet' and the device 'mps' (due to an incompatibility)
NotImplementedError: if the policy.type is 'vqbet' and the policy device 'mps' (due to an incompatibility)
Returns:
PreTrainedPolicy: _description_
@ -111,7 +108,7 @@ def make_policy(
# https://github.com/pytorch/pytorch/issues/77764. As a temporary fix, you can set the environment
# variable `PYTORCH_ENABLE_MPS_FALLBACK=1` to use the CPU as a fallback for this op. WARNING: this will be
# slower than running natively on MPS.
if cfg.type == "vqbet" and str(device) == "mps":
if cfg.type == "vqbet" and cfg.device == "mps":
raise NotImplementedError(
"Current implementation of VQBeT does not support `mps` backend. "
"Please use `cpu` or `cuda` backend."
@ -145,7 +142,7 @@ def make_policy(
# Make a fresh policy.
policy = policy_cls(**kwargs)
policy.to(device)
policy.to(cfg.device)
assert isinstance(policy, nn.Module)
# policy = torch.compile(policy, mode="reduce-overhead")

15
lerobot/common/policies/pi0/configuration_pi0.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from dataclasses import dataclass, field
from lerobot.common.optim.optimizers import AdamWConfig
@ -76,6 +90,7 @@ class PI0Config(PreTrainedConfig):
def __post_init__(self):
super().__post_init__()
# TODO(Steven): Validate device and amp? in all policy configs?
"""Input validation (not exhaustive)."""
if self.n_action_steps > self.chunk_size:
raise ValueError(

16
lerobot/common/policies/pi0/conversion_scripts/benchmark.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import torch
from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
@ -31,7 +45,7 @@ def main():
cfg = PreTrainedConfig.from_pretrained(ckpt_torch_dir)
cfg.pretrained_path = ckpt_torch_dir
policy = make_policy(cfg, device, ds_meta=dataset.meta)
policy = make_policy(cfg, ds_meta=dataset.meta)
# policy = torch.compile(policy, mode="reduce-overhead")

16
lerobot/common/policies/pi0/conversion_scripts/compare_with_jax.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import json
import pickle
from pathlib import Path
@ -87,7 +101,7 @@ def main():
cfg = PreTrainedConfig.from_pretrained(ckpt_torch_dir)
cfg.pretrained_path = ckpt_torch_dir
policy = make_policy(cfg, device, dataset_meta)
policy = make_policy(cfg, dataset_meta)
# loss_dict = policy.forward(batch, noise=noise, time=time_beta)
# loss_dict["loss"].backward()

14
lerobot/common/policies/pi0/conversion_scripts/conversion_utils.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from transformers import GemmaConfig, PaliGemmaConfig

14
lerobot/common/policies/pi0/conversion_scripts/convert_pi0_to_hf_lerobot.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Convert pi0 parameters from Jax to Pytorch

14
lerobot/common/policies/pi0/flex_attention.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import torch
import torch.nn.functional as F # noqa: N812
from packaging.version import Version

2
lerobot/common/policies/pi0/modeling_pi0.py
View File

@ -313,7 +313,7 @@ class PI0Policy(PreTrainedPolicy):
state = self.prepare_state(batch)
lang_tokens, lang_masks = self.prepare_language(batch)
actions = self.prepare_action(batch)
actions_is_pad = batch.get("actions_id_pad")
actions_is_pad = batch.get("actions_is_pad")
loss_dict = {}
losses = self.model.forward(images, img_masks, lang_tokens, lang_masks, state, actions, noise, time)

14
lerobot/common/policies/pi0/paligemma_with_expert.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import List, Optional, Union
import torch

20
lerobot/common/policies/pretrained.py
View File

@ -1,3 +1,16 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import abc
import logging
import os
@ -73,7 +86,6 @@ class PreTrainedPolicy(nn.Module, HubMixin, abc.ABC):
cache_dir: str | Path | None = None,
local_files_only: bool = False,
revision: str | None = None,
map_location: str = "cpu",
strict: bool = False,
**kwargs,
) -> T:
@ -98,7 +110,7 @@ class PreTrainedPolicy(nn.Module, HubMixin, abc.ABC):
if os.path.isdir(model_id):
print("Loading weights from local directory")
model_file = os.path.join(model_id, SAFETENSORS_SINGLE_FILE)
policy = cls._load_as_safetensor(instance, model_file, map_location, strict)
policy = cls._load_as_safetensor(instance, model_file, config.device, strict)
else:
try:
model_file = hf_hub_download(
@ -112,13 +124,13 @@ class PreTrainedPolicy(nn.Module, HubMixin, abc.ABC):
token=token,
local_files_only=local_files_only,
)
policy = cls._load_as_safetensor(instance, model_file, map_location, strict)
policy = cls._load_as_safetensor(instance, model_file, config.device, strict)
except HfHubHTTPError as e:
raise FileNotFoundError(
f"{SAFETENSORS_SINGLE_FILE} not found on the HuggingFace Hub in {model_id}"
) from e
policy.to(map_location)
policy.to(config.device)
policy.eval()
return policy

126
lerobot/common/robots/lekiwi/README.md
View File

@ -23,6 +23,9 @@ Follow this [README](https://github.com/SIGRobotics-UIUC/LeKiwi). It contains th
Before assembling, you will first need to configure your motors. To this end, we provide a nice script, so let's first install LeRobot. After configuration, we will also guide you through assembly.
### Wired version
If you have the **wired** LeKiwi version you can skip the installation of the Raspberry Pi and setting up SSH. You can also run all commands directly on your PC for both the LeKiwi scripts and the leader arm scripts for teleoperating.
## B. Install software on Pi
Now we have to setup the remote PC that will run on the LeKiwi Robot. This is normally a Raspberry Pi, but can be any PC that can run on 5V and has enough usb ports (2 or more) for the cameras and motor control board.
@ -246,6 +249,110 @@ class LeKiwiRobotConfig(RobotConfig):
}
)
teleop_keys: dict[str, str] = field(
default_factory=lambda: {
# Movement
"forward": "w",
"backward": "s",
"left": "a",
"right": "d",
"rotate_left": "z",
"rotate_right": "x",
# Speed control
"speed_up": "r",
"speed_down": "f",
# quit teleop
"quit": "q",
}
)
mock: bool = False
```
## Wired version
For the wired LeKiwi version your configured IP address should refer to your own laptop (127.0.0.1), because leader arm and LeKiwi are in this case connected to own laptop. Below and example configuration for this wired setup:
```python
@RobotConfig.register_subclass("lekiwi")
@dataclass
class LeKiwiRobotConfig(RobotConfig):
# `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
# Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
# the number of motors in your follower arms.
max_relative_target: int | None = None
# Network Configuration
ip: str = "127.0.0.1"
port: int = 5555
video_port: int = 5556
cameras: dict[str, CameraConfig] = field(
default_factory=lambda: {
"front": OpenCVCameraConfig(
camera_index=0, fps=30, width=640, height=480, rotation=90
),
"wrist": OpenCVCameraConfig(
camera_index=1, fps=30, width=640, height=480, rotation=180
),
}
)
calibration_dir: str = ".cache/calibration/lekiwi"
leader_arms: dict[str, MotorsBusConfig] = field(
default_factory=lambda: {
"main": FeetechMotorsBusConfig(
port="/dev/tty.usbmodem585A0077581",
motors={
# name: (index, model)
"shoulder_pan": [1, "sts3215"],
"shoulder_lift": [2, "sts3215"],
"elbow_flex": [3, "sts3215"],
"wrist_flex": [4, "sts3215"],
"wrist_roll": [5, "sts3215"],
"gripper": [6, "sts3215"],
},
),
}
)
follower_arms: dict[str, MotorsBusConfig] = field(
default_factory=lambda: {
"main": FeetechMotorsBusConfig(
port="/dev/tty.usbmodem58760431061",
motors={
# name: (index, model)
"shoulder_pan": [1, "sts3215"],
"shoulder_lift": [2, "sts3215"],
"elbow_flex": [3, "sts3215"],
"wrist_flex": [4, "sts3215"],
"wrist_roll": [5, "sts3215"],
"gripper": [6, "sts3215"],
"left_wheel": (7, "sts3215"),
"back_wheel": (8, "sts3215"),
"right_wheel": (9, "sts3215"),
},
),
}
)
teleop_keys: dict[str, str] = field(
default_factory=lambda: {
# Movement
"forward": "w",
"backward": "s",
"left": "a",
"right": "d",
"rotate_left": "z",
"rotate_right": "x",
# Speed control
"speed_up": "r",
"speed_down": "f",
# quit teleop
"quit": "q",
}
)
mock: bool = False
```
@ -272,6 +379,9 @@ python lerobot/scripts/control_robot.py \
--control.arms='["main_follower"]'
```
### Wired version
If you have the **wired** LeKiwi version please run all commands including this calibration command on your laptop.
### Calibrate leader arm
Then to calibrate the leader arm (which is attached to the laptop/pc). You will need to move the leader arm to these positions sequentially:
@ -326,6 +436,9 @@ You should see on your laptop something like this: ```[INFO] Connected to remote
> [!TIP]
> If you use a different keyboard you can change the keys for each command in the [`LeKiwiRobotConfig`](../lerobot/common/robot_devices/robots/configs.py).
### Wired version
If you have the **wired** LeKiwi version please run all commands including both these teleoperation commands on your laptop.
## Troubleshoot communication
If you are having trouble connecting to the Mobile SO100, follow these steps to diagnose and resolve the issue.
@ -364,6 +477,13 @@ Make sure the configuration file on both your laptop/pc and the Raspberry Pi is
# G. Record a dataset
Once you're familiar with teleoperation, you can record your first dataset with LeKiwi.
To start the program on LeKiwi, SSH into your Raspberry Pi, and run `conda activate lerobot` and this script:
```bash
python lerobot/scripts/control_robot.py \
--robot.type=lekiwi \
--control.type=remote_robot
```
If you want to use the Hugging Face hub features for uploading your dataset and you haven't previously done it, make sure you've logged in using a write-access token, which can be generated from the [Hugging Face settings](https://huggingface.co/settings/tokens):
```bash
huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
@ -374,8 +494,7 @@ Store your Hugging Face repository name in a variable to run these commands:
HF_USER=$(huggingface-cli whoami | head -n 1)
echo $HF_USER
```
Record 2 episodes and upload your dataset to the hub:
On your laptop then run this command to record 2 episodes and upload your dataset to the hub:
```bash
python lerobot/scripts/control_robot.py \
--robot.type=lekiwi \
@ -393,6 +512,9 @@ python lerobot/scripts/control_robot.py \
Note: You can resume recording by adding `--control.resume=true`.
### Wired version
If you have the **wired** LeKiwi version please run all commands including both these record dataset commands on your laptop.
# H. Visualize a dataset
If you uploaded your dataset to the hub with `--control.push_to_hub=true`, you can [visualize your dataset online](https://huggingface.co/spaces/lerobot/visualize_dataset) by copy pasting your repo id given by:

14
lerobot/common/robots/lekiwi/lekiwi_remote.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import base64
import json
import threading

14
lerobot/common/robots/manipulator.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Contains logic to instantiate a robot, read information from its motors and cameras,
and send orders to its motors.
"""

38
lerobot/common/robots/mobile_manipulator.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import base64
import json
import os
@ -393,21 +407,19 @@ class MobileManipulator:
for name in self.leader_arms:
pos = self.leader_arms[name].read("Present_Position")
pos_tensor = torch.from_numpy(pos).float()
# Instead of pos_tensor.item(), use tolist() to convert the entire tensor to a list
arm_positions.extend(pos_tensor.tolist())
# (The rest of your code for generating wheel commands remains unchanged)
x_cmd = 0.0 # m/s forward/backward
y_cmd = 0.0 # m/s lateral
y_cmd = 0.0 # m/s forward/backward
x_cmd = 0.0 # m/s lateral
theta_cmd = 0.0 # deg/s rotation
if self.pressed_keys["forward"]:
x_cmd += xy_speed
if self.pressed_keys["backward"]:
x_cmd -= xy_speed
if self.pressed_keys["left"]:
y_cmd += xy_speed
if self.pressed_keys["right"]:
if self.pressed_keys["backward"]:
y_cmd -= xy_speed
if self.pressed_keys["left"]:
x_cmd += xy_speed
if self.pressed_keys["right"]:
x_cmd -= xy_speed
if self.pressed_keys["rotate_left"]:
theta_cmd += theta_speed
if self.pressed_keys["rotate_right"]:
@ -585,8 +597,8 @@ class MobileManipulator:
# Create the body velocity vector [x, y, theta_rad].
velocity_vector = np.array([x_cmd, y_cmd, theta_rad])
# Define the wheel mounting angles with a -90° offset.
angles = np.radians(np.array([240, 120, 0]) - 90)
# Define the wheel mounting angles (defined from y axis cw)
angles = np.radians(np.array([300, 180, 60]))
# Build the kinematic matrix: each row maps body velocities to a wheel’s linear speed.
# The third column (base_radius) accounts for the effect of rotation.
m = np.array([[np.cos(a), np.sin(a), base_radius] for a in angles])
@ -642,8 +654,8 @@ class MobileManipulator:
# Compute each wheel’s linear speed (m/s) from its angular speed.
wheel_linear_speeds = wheel_radps * wheel_radius
# Define the wheel mounting angles with a -90° offset.
angles = np.radians(np.array([240, 120, 0]) - 90)
# Define the wheel mounting angles (defined from y axis cw)
angles = np.radians(np.array([300, 180, 60]))
m = np.array([[np.cos(a), np.sin(a), base_radius] for a in angles])
# Solve the inverse kinematics: body_velocity = M⁻¹ · wheel_linear_speeds.

235
lerobot/common/robots/so100/README.md
View File

@ -4,8 +4,8 @@
- [A. Source the parts](#a-source-the-parts)
- [B. Install LeRobot](#b-install-lerobot)
- [C. Configure the motors](#c-configure-the-motors)
- [D. Assemble the arms](#d-assemble-the-arms)
- [C. Configure the Motors](#c-configure-the-motors)
- [D. Step-by-Step Assembly Instructions](#d-step-by-step-assembly-instructions)
- [E. Calibrate](#e-calibrate)
- [F. Teleoperate](#f-teleoperate)
- [G. Record a dataset](#g-record-a-dataset)
@ -70,6 +70,7 @@ conda install -y -c conda-forge "opencv>=4.10.0"
```
Great :hugs:! You are now done installing LeRobot and we can begin assembling the SO100 arms :robot:.
Every time you now want to use LeRobot you can go to the `~/lerobot` folder where we installed LeRobot and run one of the commands.
## C. Configure the motors
> [!NOTE]
@ -98,22 +99,22 @@ Example output when identifying the leader arm's port (e.g., `/dev/tty.usbmodem5
```
Finding all available ports for the MotorBus.
['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
Remove the usb cable from your MotorsBus and press Enter when done.
[...Disconnect leader arm and press Enter...]
The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0031751
The port of this MotorsBus is /dev/tty.usbmodem575E0031751
Reconnect the usb cable.
```
Example output when identifying the follower arm's port (e.g., `/dev/tty.usbmodem575E0032081`, or possibly `/dev/ttyACM1` on Linux):
```
Finding all available ports for the MotorBus.
['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
Remove the usb cable from your MotorsBus and press Enter when done.
[...Disconnect follower arm and press Enter...]
The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0032081
The port of this MotorsBus is /dev/tty.usbmodem575E0032081
Reconnect the usb cable.
```
@ -221,19 +222,13 @@ Redo the process for all your motors until ID 6. Do the same for the 6 motors of
Follow the video for removing gears. You need to remove the gear for the motors of the leader arm. As a result, you will only use the position encoding of the motor and reduce friction to more easily operate the leader arm.
#### c. Add motor horn to all 12 motors
## D. Step-by-Step Assembly Instructions
<details>
<summary><strong>Video adding motor horn</strong></summary>
**Step 1: Clean Parts**
- Remove all support material from the 3D-printed parts.
---
<video src="https://github.com/user-attachments/assets/ef3391a4-ad05-4100-b2bd-1699bf86c969"></video>
</details>
Follow the video for adding the motor horn. For SO-100, you need to align the holes on the motor horn to the motor spline to be approximately 1:30, 4:30, 7:30 and 10:30.
Try to avoid rotating the motor while doing so to keep position 2048 set during configuration. It is especially tricky for the leader motors as it is more sensible without the gears, but it's ok if it's a bit rotated.
## D. Assemble the arms
### Additional Guidance
<details>
<summary><strong>Video assembling arms</strong></summary>
@ -242,7 +237,211 @@ Try to avoid rotating the motor while doing so to keep position 2048 set during
</details>
Follow the video for assembling the arms. It is important to insert the cables into the motor that is being assembled before you assemble the motor into the arm! Inserting the cables beforehand is much easier than doing this afterward. The first arm should take a bit more than 1 hour to assemble, but once you get used to it, you can do it under 1 hour for the second arm.
**Note:**
This video provides visual guidance for assembling the arms, but it doesn't specify when or how to do the wiring. Inserting the cables beforehand is much easier than doing it afterward. The first arm may take a bit more than 1 hour to assemble, but once you get used to it, you can assemble the second arm in under 1 hour.
---
### First Motor
**Step 2: Insert Wires**
- Insert two wires into the first motor.
<img src="../media/tutorial/img1.jpg" style="height:300px;">
**Step 3: Install in Base**
- Place the first motor into the base.
<img src="../media/tutorial/img2.jpg" style="height:300px;">
**Step 4: Secure Motor**
- Fasten the motor with 4 screws. Two from the bottom and two from top.
**Step 5: Attach Motor Holder**
- Slide over the first motor holder and fasten it using two screws (one on each side).
<img src="../media/tutorial/img4.jpg" style="height:300px;">
**Step 6: Attach Motor Horns**
- Install both motor horns, securing the top horn with a screw. Try not to move the motor position when attaching the motor horn, especially for the leader arms, where we removed the gears.
<img src="../media/tutorial/img5.jpg" style="height:300px;">
<details>
<summary><strong>Video adding motor horn</strong></summary>
<video src="https://github.com/user-attachments/assets/ef3391a4-ad05-4100-b2bd-1699bf86c969"></video>
</details>
**Step 7: Attach Shoulder Part**
- Route one wire to the back of the robot and the other to the left or in photo towards you (see photo).
- Attach the shoulder part.
<img src="../media/tutorial/img6.jpg" style="height:300px;">
**Step 8: Secure Shoulder**
- Tighten the shoulder part with 4 screws on top and 4 on the bottom
*(access bottom holes by turning the shoulder).*
---
### Second Motor Assembly
**Step 9: Install Motor 2**
- Slide the second motor in from the top and link the wire from motor 1 to motor 2.
<img src="../media/tutorial/img8.jpg" style="height:300px;">
**Step 10: Attach Shoulder Holder**
- Add the shoulder motor holder.
- Ensure the wire from motor 1 to motor 2 goes behind the holder while the other wire is routed upward (see photo).
- This part can be tight to assemble, you can use a workbench like the image or a similar setup to push the part around the motor.
<div style="display: flex;">
<img src="../media/tutorial/img9.jpg" style="height:250px;">
<img src="../media/tutorial/img10.jpg" style="height:250px;">
<img src="../media/tutorial/img12.jpg" style="height:250px;">
</div>
**Step 11: Secure Motor 2**
- Fasten the second motor with 4 screws.
**Step 12: Attach Motor Horn**
- Attach both motor horns to motor 2, again use the horn screw.
**Step 13: Attach Base**
- Install the base attachment using 2 screws.
<img src="../media/tutorial/img11.jpg" style="height:300px;">
**Step 14: Attach Upper Arm**
- Attach the upper arm with 4 screws on each side.
<img src="../media/tutorial/img13.jpg" style="height:300px;">
---
### Third Motor Assembly
**Step 15: Install Motor 3**
- Route the motor cable from motor 2 through the cable holder to motor 3, then secure motor 3 with 4 screws.
**Step 16: Attach Motor Horn**
- Attach both motor horns to motor 3 and secure one again with a horn screw.
<img src="../media/tutorial/img14.jpg" style="height:300px;">
**Step 17: Attach Forearm**
- Connect the forearm to motor 3 using 4 screws on each side.
<img src="../media/tutorial/img15.jpg" style="height:300px;">
---
### Fourth Motor Assembly
**Step 18: Install Motor 4**
- Slide in motor 4, attach the cable from motor 3, and secure the cable in its holder with a screw.
<div style="display: flex;">
<img src="../media/tutorial/img16.jpg" style="height:300px;">
<img src="../media/tutorial/img19.jpg" style="height:300px;">
</div>
**Step 19: Attach Motor Holder 4**
- Install the fourth motor holder (a tight fit). Ensure one wire is routed upward and the wire from motor 3 is routed downward (see photo).
<img src="../media/tutorial/img17.jpg" style="height:300px;">
**Step 20: Secure Motor 4 & Attach Horn**
- Fasten motor 4 with 4 screws and attach its motor horns, use for one a horn screw.
<img src="../media/tutorial/img18.jpg" style="height:300px;">
---
### Wrist Assembly
**Step 21: Install Motor 5**
- Insert motor 5 into the wrist holder and secure it with 2 front screws.
<img src="../media/tutorial/img20.jpg" style="height:300px;">
**Step 22: Attach Wrist**
- Connect the wire from motor 4 to motor 5. And already insert the other wire for the gripper.
- Secure the wrist to motor 4 using 4 screws on both sides.
<img src="../media/tutorial/img22.jpg" style="height:300px;">
**Step 23: Attach Wrist Horn**
- Install only one motor horn on the wrist motor and secure it with a horn screw.
<img src="../media/tutorial/img23.jpg" style="height:300px;">
---
### Follower Configuration
**Step 24: Attach Gripper**
- Attach the gripper to motor 5.
<img src="../media/tutorial/img24.jpg" style="height:300px;">
**Step 25: Install Gripper Motor**
- Insert the gripper motor, connect the motor wire from motor 5 to motor 6, and secure it with 3 screws on each side.
<img src="../media/tutorial/img25.jpg" style="height:300px;">
**Step 26: Attach Gripper Horn & Claw**
- Attach the motor horns and again use a horn screw.
- Install the gripper claw and secure it with 4 screws on both sides.
<img src="../media/tutorial/img26.jpg" style="height:300px;">
**Step 27: Mount Controller**
- Attach the motor controller on the back.
<div style="display: flex;">
<img src="../media/tutorial/img27.jpg" style="height:300px;">
<img src="../media/tutorial/img28.jpg" style="height:300px;">
</div>
*Assembly complete – proceed to Leader arm assembly.*
---
### Leader Configuration
For the leader configuration, perform **Steps 1–23**. Make sure that you removed the motor gears from the motors.
**Step 24: Attach Leader Holder**
- Mount the leader holder onto the wrist and secure it with a screw.
<img src="../media/tutorial/img29.jpg" style="height:300px;">
**Step 25: Attach Handle**
- Attach the handle to motor 5 using 4 screws.
<img src="../media/tutorial/img30.jpg" style="height:300px;">
**Step 26: Install Gripper Motor**
- Insert the gripper motor, secure it with 3 screws on each side, attach a motor horn using a horn screw, and connect the motor wire.
<img src="../media/tutorial/img31.jpg" style="height:300px;">
**Step 27: Attach Trigger**
- Attach the follower trigger with 4 screws.
<img src="../media/tutorial/img32.jpg" style="height:300px;">
**Step 28: Mount Controller**
- Attach the motor controller on the back.
<div style="display: flex;">
<img src="../media/tutorial/img27.jpg" style="height:300px;">
<img src="../media/tutorial/img28.jpg" style="height:300px;">
</div>
*Assembly complete – proceed to calibration.*
## E. Calibrate

30
lerobot/common/utils/control_utils.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
########################################################################################
# Utilities
########################################################################################
@ -18,6 +32,7 @@ from termcolor import colored
from lerobot.common.datasets.image_writer import safe_stop_image_writer
from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
from lerobot.common.datasets.utils import get_features_from_robot
from lerobot.common.policies.pretrained import PreTrainedPolicy
from lerobot.common.robots.utils import Robot
from lerobot.common.utils.robot_utils import busy_wait
from lerobot.common.utils.utils import get_safe_torch_device, has_method
@ -179,8 +194,6 @@ def record_episode(
episode_time_s,
display_cameras,
policy,
device,
use_amp,
fps,
single_task,
):
@ -191,8 +204,6 @@ def record_episode(
dataset=dataset,
events=events,
policy=policy,
device=device,
use_amp=use_amp,
fps=fps,
teleoperate=policy is None,
single_task=single_task,
@ -207,9 +218,7 @@ def control_loop(
display_cameras=False,
dataset: LeRobotDataset | None = None,
events=None,
policy=None,
device: torch.device | str | None = None,
use_amp: bool | None = None,
policy: PreTrainedPolicy = None,
fps: int | None = None,
single_task: str | None = None,
):
@ -232,9 +241,6 @@ def control_loop(
if dataset is not None and fps is not None and dataset.fps != fps:
raise ValueError(f"The dataset fps should be equal to requested fps ({dataset['fps']} != {fps}).")
if isinstance(device, str):
device = get_safe_torch_device(device)
timestamp = 0
start_episode_t = time.perf_counter()
while timestamp < control_time_s:
@ -246,7 +252,9 @@ def control_loop(
observation = robot.capture_observation()
if policy is not None:
pred_action = predict_action(observation, policy, device, use_amp)
pred_action = predict_action(
observation, policy, get_safe_torch_device(policy.config.device), policy.config.use_amp
)
# Action can eventually be clipped using `max_relative_target`,
# so action actually sent is saved in the dataset.
action = robot.send_action(pred_action)

14
lerobot/common/utils/hub.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from pathlib import Path
from tempfile import TemporaryDirectory
from typing import Any, Type, TypeVar

14
lerobot/common/utils/robot_utils.py
View File

@ -1,3 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import platform
import time

46
lerobot/common/utils/utils.py
View File

@ -17,6 +17,7 @@ import logging
import os
import os.path as osp
import platform
import subprocess
from copy import copy
from datetime import datetime, timezone
from pathlib import Path
@ -50,8 +51,10 @@ def auto_select_torch_device() -> torch.device:
return torch.device("cpu")
# TODO(Steven): Remove log. log shouldn't be an argument, this should be handled by the logger level
def get_safe_torch_device(try_device: str, log: bool = False) -> torch.device:
"""Given a string, return a torch.device with checks on whether the device is available."""
try_device = str(try_device)
match try_device:
case "cuda":
assert torch.cuda.is_available()
@ -84,6 +87,7 @@ def get_safe_dtype(dtype: torch.dtype, device: str | torch.device):
def is_torch_device_available(try_device: str) -> bool:
try_device = str(try_device) # Ensure try_device is a string
if try_device == "cuda":
return torch.cuda.is_available()
elif try_device == "mps":
@ -91,7 +95,7 @@ def is_torch_device_available(try_device: str) -> bool:
elif try_device == "cpu":
return True
else:
raise ValueError(f"Unknown device '{try_device}.")
raise ValueError(f"Unknown device {try_device}. Supported devices are: cuda, mps or cpu.")
def is_amp_available(device: str):
@ -165,23 +169,31 @@ def capture_timestamp_utc():
def say(text, blocking=False):
# Check if mac, linux, or windows.
if platform.system() == "Darwin":
cmd = f'say "{text}"'
if not blocking:
cmd += " &"
elif platform.system() == "Linux":
cmd = f'spd-say "{text}"'
if blocking:
cmd += " --wait"
elif platform.system() == "Windows":
# TODO(rcadene): Make blocking option work for Windows
cmd = (
'PowerShell -Command "Add-Type -AssemblyName System.Speech; '
f"(New-Object System.Speech.Synthesis.SpeechSynthesizer).Speak('{text}')\""
)
system = platform.system()
os.system(cmd)
if system == "Darwin":
cmd = ["say", text]
elif system == "Linux":
cmd = ["spd-say", text]
if blocking:
cmd.append("--wait")
elif system == "Windows":
cmd = [
"PowerShell",
"-Command",
"Add-Type -AssemblyName System.Speech; "
f"(New-Object System.Speech.Synthesis.SpeechSynthesizer).Speak('{text}')",
]
else:
raise RuntimeError("Unsupported operating system for text-to-speech.")
if blocking:
subprocess.run(cmd, check=True)
else:
subprocess.Popen(cmd, creationflags=subprocess.CREATE_NO_WINDOW if system == "Windows" else 0)
def log_say(text, play_sounds, blocking=False):

43
lerobot/configs/control.py
View File

@ -1,14 +1,25 @@
import logging
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from dataclasses import dataclass
from pathlib import Path
import draccus
from lerobot.common.robots import RobotConfig
from lerobot.common.utils.utils import auto_select_torch_device, is_amp_available, is_torch_device_available
from lerobot.configs import parser
from lerobot.configs.policies import PreTrainedConfig
from lerobot.configs.train import TrainPipelineConfig
@dataclass
@ -43,11 +54,6 @@ class RecordControlConfig(ControlConfig):
# Root directory where the dataset will be stored (e.g. 'dataset/path').
root: str | Path | None = None
policy: PreTrainedConfig | None = None
# TODO(rcadene, aliberts): By default, use device and use_amp values from policy checkpoint.
device: str | None = None # cuda | cpu | mps
# `use_amp` determines whether to use Automatic Mixed Precision (AMP) for training and evaluation. With AMP,
# automatic gradient scaling is used.
use_amp: bool | None = None
# Limit the frames per second. By default, uses the policy fps.
fps: int | None = None
# Number of seconds before starting data collection. It allows the robot devices to warmup and synchronize.
@ -90,27 +96,6 @@ class RecordControlConfig(ControlConfig):
self.policy = PreTrainedConfig.from_pretrained(policy_path, cli_overrides=cli_overrides)
self.policy.pretrained_path = policy_path
# When no device or use_amp are given, use the one from training config.
if self.device is None or self.use_amp is None:
train_cfg = TrainPipelineConfig.from_pretrained(policy_path)
if self.device is None:
self.device = train_cfg.device
if self.use_amp is None:
self.use_amp = train_cfg.use_amp
# Automatically switch to available device if necessary
if not is_torch_device_available(self.device):
auto_device = auto_select_torch_device()
logging.warning(f"Device '{self.device}' is not available. Switching to '{auto_device}'.")
self.device = auto_device
# Automatically deactivate AMP if necessary
if self.use_amp and not is_amp_available(self.device):
logging.warning(
f"Automatic Mixed Precision (amp) is not available on device '{self.device}'. Deactivating AMP."
)
self.use_amp = False
@ControlConfig.register_subclass("replay")
@dataclass

47
lerobot/configs/eval.py
View File

@ -1,14 +1,26 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import datetime as dt
import logging
from dataclasses import dataclass, field
from pathlib import Path
from lerobot.common import envs, policies # noqa: F401
from lerobot.common.utils.utils import auto_select_torch_device, is_amp_available, is_torch_device_available
from lerobot.configs import parser
from lerobot.configs.default import EvalConfig
from lerobot.configs.policies import PreTrainedConfig
from lerobot.configs.train import TrainPipelineConfig
@dataclass
@ -21,11 +33,6 @@ class EvalPipelineConfig:
policy: PreTrainedConfig | None = None
output_dir: Path | None = None
job_name: str | None = None
# TODO(rcadene, aliberts): By default, use device and use_amp values from policy checkpoint.
device: str | None = None # cuda | cpu | mps
# `use_amp` determines whether to use Automatic Mixed Precision (AMP) for training and evaluation. With AMP,
# automatic gradient scaling is used.
use_amp: bool = False
seed: int | None = 1000
def __post_init__(self):
@ -36,27 +43,6 @@ class EvalPipelineConfig:
self.policy = PreTrainedConfig.from_pretrained(policy_path, cli_overrides=cli_overrides)
self.policy.pretrained_path = policy_path
# When no device or use_amp are given, use the one from training config.
if self.device is None or self.use_amp is None:
train_cfg = TrainPipelineConfig.from_pretrained(policy_path)
if self.device is None:
self.device = train_cfg.device
if self.use_amp is None:
self.use_amp = train_cfg.use_amp
# Automatically switch to available device if necessary
if not is_torch_device_available(self.device):
auto_device = auto_select_torch_device()
logging.warning(f"Device '{self.device}' is not available. Switching to '{auto_device}'.")
self.device = auto_device
# Automatically deactivate AMP if necessary
if self.use_amp and not is_amp_available(self.device):
logging.warning(
f"Automatic Mixed Precision (amp) is not available on device '{self.device}'. Deactivating AMP."
)
self.use_amp = False
else:
logging.warning(
"No pretrained path was provided, evaluated policy will be built from scratch (random weights)."
@ -73,11 +59,6 @@ class EvalPipelineConfig:
eval_dir = f"{now:%Y-%m-%d}/{now:%H-%M-%S}_{self.job_name}"
self.output_dir = Path("outputs/eval") / eval_dir
if self.device is None:
raise ValueError("Set one of the following device: cuda, cpu or mps")
elif self.device == "cuda" and self.use_amp is None:
raise ValueError("Set 'use_amp' to True or False.")
@classmethod
def __get_path_fields__(cls) -> list[str]:
"""This enables the parser to load config from the policy using `--policy.path=local/dir`"""

109
lerobot/configs/parser.py
View File

@ -1,4 +1,19 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import importlib
import inspect
import pkgutil
import sys
from argparse import ArgumentError
from functools import wraps
@ -10,6 +25,7 @@ import draccus
from lerobot.common.utils.utils import has_method
PATH_KEY = "path"
PLUGIN_DISCOVERY_SUFFIX = "discover_packages_path"
draccus.set_config_type("json")
@ -45,6 +61,86 @@ def parse_arg(arg_name: str, args: Sequence[str] | None = None) -> str | None:
return None
def parse_plugin_args(plugin_arg_suffix: str, args: Sequence[str]) -> dict:
"""Parse plugin-related arguments from command-line arguments.
This function extracts arguments from command-line arguments that match a specified suffix pattern.
It processes arguments in the format '--key=value' and returns them as a dictionary.
Args:
plugin_arg_suffix (str): The suffix to identify plugin-related arguments.
cli_args (Sequence[str]): A sequence of command-line arguments to parse.
Returns:
dict: A dictionary containing the parsed plugin arguments where:
- Keys are the argument names (with '--' prefix removed if present)
- Values are the corresponding argument values
Example:
>>> args = ['--env.discover_packages_path=my_package',
... '--other_arg=value']
>>> parse_plugin_args('discover_packages_path', args)
{'env.discover_packages_path': 'my_package'}
"""
plugin_args = {}
for arg in args:
if "=" in arg and plugin_arg_suffix in arg:
key, value = arg.split("=", 1)
# Remove leading '--' if present
if key.startswith("--"):
key = key[2:]
plugin_args[key] = value
return plugin_args
class PluginLoadError(Exception):
"""Raised when a plugin fails to load."""
def load_plugin(plugin_path: str) -> None:
"""Load and initialize a plugin from a given Python package path.
This function attempts to load a plugin by importing its package and any submodules.
Plugin registration is expected to happen during package initialization, i.e. when
the package is imported the gym environment should be registered and the config classes
registered with their parents using the `register_subclass` decorator.
Args:
plugin_path (str): The Python package path to the plugin (e.g. "mypackage.plugins.myplugin")
Raises:
PluginLoadError: If the plugin cannot be loaded due to import errors or if the package path is invalid.
Examples:
>>> load_plugin("external_plugin.core") # Loads plugin from external package
Notes:
- The plugin package should handle its own registration during import
- All submodules in the plugin package will be imported
- Implementation follows the plugin discovery pattern from Python packaging guidelines
See Also:
https://packaging.python.org/en/latest/guides/creating-and-discovering-plugins/
"""
try:
package_module = importlib.import_module(plugin_path, __package__)
except (ImportError, ModuleNotFoundError) as e:
raise PluginLoadError(
f"Failed to load plugin '{plugin_path}'. Verify the path and installation: {str(e)}"
) from e
def iter_namespace(ns_pkg):
return pkgutil.iter_modules(ns_pkg.__path__, ns_pkg.__name__ + ".")
try:
for _finder, pkg_name, _ispkg in iter_namespace(package_module):
importlib.import_module(pkg_name)
except ImportError as e:
raise PluginLoadError(
f"Failed to load plugin '{plugin_path}'. Verify the path and installation: {str(e)}"
) from e
def get_path_arg(field_name: str, args: Sequence[str] | None = None) -> str | None:
return parse_arg(f"{field_name}.{PATH_KEY}", args)
@ -92,10 +188,13 @@ def filter_path_args(fields_to_filter: str | list[str], args: Sequence[str] | No
def wrap(config_path: Path | None = None):
"""
HACK: Similar to draccus.wrap but does two additional things:
HACK: Similar to draccus.wrap but does three additional things:
- Will remove '.path' arguments from CLI in order to process them later on.
- If a 'config_path' is passed and the main config class has a 'from_pretrained' method, will
initialize it from there to allow to fetch configs from the hub directly
- Will load plugins specified in the CLI arguments. These plugins will typically register
their own subclasses of config classes, so that draccus can find the right class to instantiate
from the CLI '.type' arguments
"""
def wrapper_outer(fn):
@ -108,6 +207,14 @@ def wrap(config_path: Path | None = None):
args = args[1:]
else:
cli_args = sys.argv[1:]
plugin_args = parse_plugin_args(PLUGIN_DISCOVERY_SUFFIX, cli_args)
for plugin_cli_arg, plugin_path in plugin_args.items():
try:
load_plugin(plugin_path)
except PluginLoadError as e:
# add the relevant CLI arg to the error message
raise PluginLoadError(f"{e}\nFailed plugin CLI Arg: {plugin_cli_arg}") from e
cli_args = filter_arg(plugin_cli_arg, cli_args)
config_path_cli = parse_arg("config_path", cli_args)
if has_method(argtype, "__get_path_fields__"):
path_fields = argtype.__get_path_fields__()

31
lerobot/configs/policies.py
View File

@ -1,4 +1,18 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import abc
import logging
import os
from dataclasses import dataclass, field
from pathlib import Path
@ -12,6 +26,7 @@ from huggingface_hub.errors import HfHubHTTPError
from lerobot.common.optim.optimizers import OptimizerConfig
from lerobot.common.optim.schedulers import LRSchedulerConfig
from lerobot.common.utils.hub import HubMixin
from lerobot.common.utils.utils import auto_select_torch_device, is_amp_available, is_torch_device_available
from lerobot.configs.types import FeatureType, NormalizationMode, PolicyFeature
# Generic variable that is either PreTrainedConfig or a subclass thereof
@ -40,8 +55,24 @@ class PreTrainedConfig(draccus.ChoiceRegistry, HubMixin, abc.ABC):
input_features: dict[str, PolicyFeature] = field(default_factory=dict)
output_features: dict[str, PolicyFeature] = field(default_factory=dict)
device: str | None = None # cuda | cpu | mp
# `use_amp` determines whether to use Automatic Mixed Precision (AMP) for training and evaluation. With AMP,
# automatic gradient scaling is used.
use_amp: bool = False
def __post_init__(self):
self.pretrained_path = None
if not self.device or not is_torch_device_available(self.device):
auto_device = auto_select_torch_device()
logging.warning(f"Device '{self.device}' is not available. Switching to '{auto_device}'.")
self.device = auto_device.type
# Automatically deactivate AMP if necessary
if self.use_amp and not is_amp_available(self.device):
logging.warning(
f"Automatic Mixed Precision (amp) is not available on device '{self.device}'. Deactivating AMP."
)
self.use_amp = False
@property
def type(self) -> str:

31
lerobot/configs/train.py
View File

@ -1,5 +1,17 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import datetime as dt
import logging
import os
from dataclasses import dataclass, field
from pathlib import Path
@ -13,7 +25,6 @@ from lerobot.common import envs
from lerobot.common.optim import OptimizerConfig
from lerobot.common.optim.schedulers import LRSchedulerConfig
from lerobot.common.utils.hub import HubMixin
from lerobot.common.utils.utils import auto_select_torch_device, is_amp_available
from lerobot.configs import parser
from lerobot.configs.default import DatasetConfig, EvalConfig, WandBConfig
from lerobot.configs.policies import PreTrainedConfig
@ -35,10 +46,6 @@ class TrainPipelineConfig(HubMixin):
# Note that when resuming a run, the default behavior is to use the configuration from the checkpoint,
# regardless of what's provided with the training command at the time of resumption.
resume: bool = False
device: str | None = None # cuda | cpu | mp
# `use_amp` determines whether to use Automatic Mixed Precision (AMP) for training and evaluation. With AMP,
# automatic gradient scaling is used.
use_amp: bool = False
# `seed` is used for training (eg: model initialization, dataset shuffling)
# AND for the evaluation environments.
seed: int | None = 1000
@ -61,18 +68,6 @@ class TrainPipelineConfig(HubMixin):
self.checkpoint_path = None
def validate(self):
if not self.device:
logging.warning("No device specified, trying to infer device automatically")
device = auto_select_torch_device()
self.device = device.type
# Automatically deactivate AMP if necessary
if self.use_amp and not is_amp_available(self.device):
logging.warning(
f"Automatic Mixed Precision (amp) is not available on device '{self.device}'. Deactivating AMP."
)
self.use_amp = False
# HACK: We parse again the cli args here to get the pretrained paths if there was some.
policy_path = parser.get_path_arg("policy")
if policy_path:

13
lerobot/configs/types.py
View File

@ -1,3 +1,16 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Note: We subclass str so that serialization is straightforward
# https://stackoverflow.com/questions/24481852/serialising-an-enum-member-to-json
from dataclasses import dataclass

13
lerobot/scripts/configure_motor.py
View File

@ -1,3 +1,16 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This script configure a single motor at a time to a given ID and baudrate.

17
lerobot/scripts/control_robot.py
View File

@ -1,3 +1,16 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Utilities to control a robot.
@ -254,7 +267,7 @@ def record(
)
# Load pretrained policy
policy = None if cfg.policy is None else make_policy(cfg.policy, cfg.device, ds_meta=dataset.meta)
policy = None if cfg.policy is None else make_policy(cfg.policy, ds_meta=dataset.meta)
if not robot.is_connected:
robot.connect()
@ -285,8 +298,6 @@ def record(
episode_time_s=cfg.episode_time_s,
display_cameras=cfg.display_cameras,
policy=policy,
device=cfg.device,
use_amp=cfg.use_amp,
fps=cfg.fps,
single_task=cfg.single_task,
)

13
lerobot/scripts/control_sim_robot.py
View File

@ -1,3 +1,16 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Utilities to control a robot in simulation.

10
lerobot/scripts/eval.py
View File

@ -454,11 +454,11 @@ def _compile_episode_data(
@parser.wrap()
def eval(cfg: EvalPipelineConfig):
def eval_main(cfg: EvalPipelineConfig):
logging.info(pformat(asdict(cfg)))
# Check device is available
device = get_safe_torch_device(cfg.device, log=True)
device = get_safe_torch_device(cfg.policy.device, log=True)
torch.backends.cudnn.benchmark = True
torch.backends.cuda.matmul.allow_tf32 = True
@ -470,14 +470,14 @@ def eval(cfg: EvalPipelineConfig):
env = make_env(cfg.env, n_envs=cfg.eval.batch_size, use_async_envs=cfg.eval.use_async_envs)
logging.info("Making policy.")
policy = make_policy(
cfg=cfg.policy,
device=device,
env_cfg=cfg.env,
)
policy.eval()
with torch.no_grad(), torch.autocast(device_type=device.type) if cfg.use_amp else nullcontext():
with torch.no_grad(), torch.autocast(device_type=device.type) if cfg.policy.use_amp else nullcontext():
info = eval_policy(
env,
policy,
@ -499,4 +499,4 @@ def eval(cfg: EvalPipelineConfig):
if __name__ == "__main__":
init_logging()
eval()
eval_main()

13
lerobot/scripts/find_motors_bus_port.py
View File

@ -1,3 +1,16 @@
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import time
from pathlib import Path

12
lerobot/scripts/train.py
View File

@ -120,7 +120,7 @@ def train(cfg: TrainPipelineConfig):
set_seed(cfg.seed)
# Check device is available
device = get_safe_torch_device(cfg.device, log=True)
device = get_safe_torch_device(cfg.policy.device, log=True)
torch.backends.cudnn.benchmark = True
torch.backends.cuda.matmul.allow_tf32 = True
@ -138,13 +138,12 @@ def train(cfg: TrainPipelineConfig):
logging.info("Creating policy")
policy = make_policy(
cfg=cfg.policy,
device=device,
ds_meta=dataset.meta,
)
logging.info("Creating optimizer and scheduler")
optimizer, lr_scheduler = make_optimizer_and_scheduler(cfg, policy)
grad_scaler = GradScaler(device, enabled=cfg.use_amp)
grad_scaler = GradScaler(device.type, enabled=cfg.policy.use_amp)
step = 0 # number of policy updates (forward + backward + optim)
@ -218,7 +217,7 @@ def train(cfg: TrainPipelineConfig):
cfg.optimizer.grad_clip_norm,
grad_scaler=grad_scaler,
lr_scheduler=lr_scheduler,
use_amp=cfg.use_amp,
use_amp=cfg.policy.use_amp,
)
# Note: eval and checkpoint happens *after* the `step`th training update has completed, so we
@ -249,7 +248,10 @@ def train(cfg: TrainPipelineConfig):
if cfg.env and is_eval_step:
step_id = get_step_identifier(step, cfg.steps)
logging.info(f"Eval policy at step {step}")
with torch.no_grad(), torch.autocast(device_type=device.type) if cfg.use_amp else nullcontext():
with (
torch.no_grad(),
torch.autocast(device_type=device.type) if cfg.policy.use_amp else nullcontext(),
):
eval_info = eval_policy(
eval_env,
policy,

26
lerobot/scripts/visualize_dataset_html.py
View File

@ -158,7 +158,7 @@ def run_server(
if major_version < 2:
return "Make sure to convert your LeRobotDataset to v2 & above."
episode_data_csv_str, columns = get_episode_data(dataset, episode_id)
episode_data_csv_str, columns, ignored_columns = get_episode_data(dataset, episode_id)
dataset_info = {
"repo_id": f"{dataset_namespace}/{dataset_name}",
"num_samples": dataset.num_frames
@ -194,7 +194,7 @@ def run_server(
]
response = requests.get(
f"https://huggingface.co/datasets/{repo_id}/resolve/main/meta/episodes.jsonl"
f"https://huggingface.co/datasets/{repo_id}/resolve/main/meta/episodes.jsonl", timeout=5
)
response.raise_for_status()
# Split into lines and parse each line as JSON
@ -218,6 +218,7 @@ def run_server(
videos_info=videos_info,
episode_data_csv_str=episode_data_csv_str,
columns=columns,
ignored_columns=ignored_columns,
)
app.run(host=host, port=port)
@ -233,9 +234,17 @@ def get_episode_data(dataset: LeRobotDataset | IterableNamespace, episode_index)
This file will be loaded by Dygraph javascript to plot data in real time."""
columns = []
selected_columns = [col for col, ft in dataset.features.items() if ft["dtype"] == "float32"]
selected_columns = [col for col, ft in dataset.features.items() if ft["dtype"] in ["float32", "int32"]]
selected_columns.remove("timestamp")
ignored_columns = []
for column_name in selected_columns:
shape = dataset.features[column_name]["shape"]
shape_dim = len(shape)
if shape_dim > 1:
selected_columns.remove(column_name)
ignored_columns.append(column_name)
# init header of csv with state and action names
header = ["timestamp"]
@ -245,16 +254,17 @@ def get_episode_data(dataset: LeRobotDataset | IterableNamespace, episode_index)
if isinstance(dataset, LeRobotDataset)
else dataset.features[column_name].shape[0]
)
header += [f"{column_name}_{i}" for i in range(dim_state)]
if "names" in dataset.features[column_name] and dataset.features[column_name]["names"]:
column_names = dataset.features[column_name]["names"]
while not isinstance(column_names, list):
column_names = list(column_names.values())[0]
else:
column_names = [f"motor_{i}" for i in range(dim_state)]
column_names = [f"{column_name}_{i}" for i in range(dim_state)]
columns.append({"key": column_name, "value": column_names})
header += column_names
selected_columns.insert(0, "timestamp")
if isinstance(dataset, LeRobotDataset):
@ -290,7 +300,7 @@ def get_episode_data(dataset: LeRobotDataset | IterableNamespace, episode_index)
csv_writer.writerows(rows)
csv_string = csv_buffer.getvalue()
return csv_string, columns
return csv_string, columns, ignored_columns
def get_episode_video_paths(dataset: LeRobotDataset, ep_index: int) -> list[str]:
@ -317,7 +327,9 @@ def get_episode_language_instruction(dataset: LeRobotDataset, ep_index: int) ->
def get_dataset_info(repo_id: str) -> IterableNamespace:
response = requests.get(f"https://huggingface.co/datasets/{repo_id}/resolve/main/meta/info.json")
response = requests.get(
f"https://huggingface.co/datasets/{repo_id}/resolve/main/meta/info.json", timeout=5
)
response.raise_for_status() # Raises an HTTPError for bad responses
dataset_info = response.json()
dataset_info["repo_id"] = repo_id

152
lerobot/templates/visualize_dataset_template.html
View File

@ -14,21 +14,7 @@
<!-- Use [Alpin.js](https://alpinejs.dev), a lightweight and easy to learn JS framework -->
<!-- Use [tailwindcss](https://tailwindcss.com/), CSS classes for styling html -->
<!-- Use [dygraphs](https://dygraphs.com/), a lightweight JS charting library -->
<body class="flex flex-col md:flex-row h-screen max-h-screen bg-slate-950 text-gray-200" x-data="createAlpineData()" @keydown.window="(e) => {
// Use the space bar to play and pause, instead of default action (e.g. scrolling)
const { keyCode, key } = e;
if (keyCode === 32 || key === ' ') {
e.preventDefault();
$refs.btnPause.classList.contains('hidden') ? $refs.btnPlay.click() : $refs.btnPause.click();
}else if (key === 'ArrowDown' || key === 'ArrowUp'){
const nextEpisodeId = key === 'ArrowDown' ? {{ episode_id }} + 1 : {{ episode_id }} - 1;
const lowestEpisodeId = {{ episodes }}.at(0);
const highestEpisodeId = {{ episodes }}.at(-1);
if(nextEpisodeId >= lowestEpisodeId && nextEpisodeId <= highestEpisodeId){
window.location.href = `./episode_${nextEpisodeId}`;
}
}
}">
<body class="flex flex-col md:flex-row h-screen max-h-screen bg-slate-950 text-gray-200" x-data="createAlpineData()">
<!-- Sidebar -->
<div x-ref="sidebar" class="bg-slate-900 p-5 break-words overflow-y-auto shrink-0 md:shrink md:w-60 md:max-h-screen">
<a href="https://github.com/huggingface/lerobot" target="_blank" class="hidden md:block">
@ -52,25 +38,55 @@
<p>Episodes:</p>
<!-- episodes menu for medium & large screens -->
<ul class="ml-2 hidden md:block">
{% for episode in episodes %}
<div class="ml-2 hidden md:block" x-data="episodePagination">
<ul>
<template x-for="episode in paginatedEpisodes" :key="episode">
<li class="font-mono text-sm mt-0.5">
<a href="episode_{{ episode }}" class="underline {% if episode_id == episode %}font-bold -ml-1{% endif %}">
Episode {{ episode }}
</a>
<a :href="'episode_' + episode"
:class="{'underline': true, 'font-bold -ml-1': episode == {{ episode_id }}}"
x-text="'Episode ' + episode"></a>
</li>
{% endfor %}
</template>
</ul>
<div class="flex items-center mt-3 text-xs" x-show="totalPages > 1">
<button @click="prevPage()"
class="px-2 py-1 bg-slate-800 rounded mr-2"
:class="{'opacity-50 cursor-not-allowed': page === 1}"
:disabled="page === 1">
&laquo; Prev
</button>
<span class="font-mono mr-2" x-text="` ${page} / ${totalPages}`"></span>
<button @click="nextPage()"
class="px-2 py-1 bg-slate-800 rounded"
:class="{'opacity-50 cursor-not-allowed': page === totalPages}"
:disabled="page === totalPages">
Next &raquo;
</button>
</div>
</div>
<!-- episodes menu for small screens -->
<div class="flex overflow-x-auto md:hidden">
{% for episode in episodes %}
<p class="font-mono text-sm mt-0.5 border-r last:border-r-0 px-2 {% if episode_id == episode %}font-bold{% endif %}">
<a href="episode_{{ episode }}" class="">
{{ episode }}
</a>
<div class="flex overflow-x-auto md:hidden" x-data="episodePagination">
<button @click="prevPage()"
class="px-2 bg-slate-800 rounded mr-2"
:class="{'opacity-50 cursor-not-allowed': page === 1}"
:disabled="page === 1">&laquo;</button>
<div class="flex">
<template x-for="(episode, index) in paginatedEpisodes" :key="episode">
<p class="font-mono text-sm mt-0.5 px-2"
:class="{
'font-bold': episode == {{ episode_id }},
'border-r': index !== paginatedEpisodes.length - 1
}">
<a :href="'episode_' + episode" x-text="episode"></a>
</p>
{% endfor %}
</template>
</div>
<button @click="nextPage()"
class="px-2 bg-slate-800 rounded ml-2"
:class="{'opacity-50 cursor-not-allowed': page === totalPages}"
:disabled="page === totalPages">&raquo; </button>
</div>
</div>
@ -208,6 +224,7 @@
</p>
</div>
<div>
<table class="text-sm border-collapse border border-slate-700" x-show="currentFrameData">
<thead>
<tr>
@ -230,14 +247,16 @@
<div class="flex gap-x-2 max-w-64 font-semibold px-1 break-all">
<input type="checkbox" :checked="isRowChecked(rowIndex)"
@change="toggleRow(rowIndex)">
<p x-text="`${rowLabels[rowIndex]}`"></p>
</div>
</td>
<template x-for="(cell, colIndex) in row">
<td x-show="cell" class="border border-slate-700">
<div class="flex gap-x-2 w-24 justify-between px-2" :class="{ 'hidden': cell.isNull }">
<div class="flex gap-x-2 justify-between px-2" :class="{ 'hidden': cell.isNull }">
<div class="flex gap-x-2">
<input type="checkbox" x-model="cell.checked" @change="updateTableValues()">
<span x-text="`${!cell.isNull ? cell.value.toFixed(2) : null}`"
<span x-text="`${!cell.isNull ? cell.label : null}`"></span>
</div>
<span class="w-14 text-right" x-text="`${!cell.isNull ? (typeof cell.value === 'number' ? cell.value.toFixed(2) : cell.value) : null}`"
:style="`color: ${cell.color}`"></span>
</div>
</td>
@ -249,6 +268,14 @@
<div id="labels" class="hidden">
</div>
{% if ignored_columns|length > 0 %}
<div class="m-2 text-orange-700 max-w-96">
Columns {{ ignored_columns }} are NOT shown since the visualizer currently does not support 2D or 3D data.
</div>
{% endif %}
</div>
</div>
</div>
@ -279,7 +306,6 @@
videosKeys: {{ videos_info | map(attribute='filename') | list | tojson }},
videosKeysSelected: [],
columns: {{ columns | tojson }},
rowLabels: {{ columns | tojson }}.reduce((colA, colB) => colA.value.length > colB.value.length ? colA : colB).value,
// alpine initialization
init() {
@ -452,6 +478,68 @@
}
};
}
document.addEventListener('alpine:init', () => {
// Episode pagination component
Alpine.data('episodePagination', () => ({
episodes: {{ episodes }},
pageSize: 100,
page: 1,
init() {
// Find which page contains the current episode_id
const currentEpisodeId = {{ episode_id }};
const episodeIndex = this.episodes.indexOf(currentEpisodeId);
if (episodeIndex !== -1) {
this.page = Math.floor(episodeIndex / this.pageSize) + 1;
}
},
get totalPages() {
return Math.ceil(this.episodes.length / this.pageSize);
},
get paginatedEpisodes() {
const start = (this.page - 1) * this.pageSize;
const end = start + this.pageSize;
return this.episodes.slice(start, end);
},
nextPage() {
if (this.page < this.totalPages) {
this.page++;
}
},
prevPage() {
if (this.page > 1) {
this.page--;
}
}
}));
});
</script>
<script>
window.addEventListener('keydown', (e) => {
// Use the space bar to play and pause, instead of default action (e.g. scrolling)
const { keyCode, key } = e;
if (keyCode === 32 || key === ' ') {
e.preventDefault();
const btnPause = document.querySelector('[x-ref="btnPause"]');
const btnPlay = document.querySelector('[x-ref="btnPlay"]');
btnPause.classList.contains('hidden') ? btnPlay.click() : btnPause.click();
} else if (key === 'ArrowDown' || key === 'ArrowUp') {
const episodes = {{ episodes }}; // Access episodes directly from the Jinja template
const nextEpisodeId = key === 'ArrowDown' ? {{ episode_id }} + 1 : {{ episode_id }} - 1;
const lowestEpisodeId = episodes.at(0);
const highestEpisodeId = episodes.at(-1);
if (nextEpisodeId >= lowestEpisodeId && nextEpisodeId <= highestEpisodeId) {
window.location.href = `./episode_${nextEpisodeId}`;
}
}
});
</script>
</body>

BIN
media/lekiwi/kiwi.webp Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 219 KiB

BIN
media/tutorial/img1.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 66 KiB

BIN
media/tutorial/img10.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 127 KiB

BIN
media/tutorial/img11.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 109 KiB

BIN
media/tutorial/img12.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 80 KiB

BIN
media/tutorial/img13.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 88 KiB

BIN
media/tutorial/img14.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 86 KiB

BIN
media/tutorial/img15.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 96 KiB

BIN
media/tutorial/img16.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 84 KiB

BIN
media/tutorial/img17.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 72 KiB

BIN
media/tutorial/img18.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 78 KiB

BIN
media/tutorial/img19.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 97 KiB

BIN
media/tutorial/img2.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 88 KiB

BIN
media/tutorial/img20.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 42 KiB

BIN
media/tutorial/img21.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 85 KiB

BIN
media/tutorial/img22.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 62 KiB

BIN
media/tutorial/img23.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 54 KiB

BIN
media/tutorial/img24.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 61 KiB

BIN
media/tutorial/img25.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 76 KiB

BIN
media/tutorial/img26.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 80 KiB

BIN
media/tutorial/img27.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 48 KiB

BIN
media/tutorial/img28.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 91 KiB

BIN
media/tutorial/img29.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 54 KiB

BIN
media/tutorial/img3.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 86 KiB

BIN
media/tutorial/img30.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 57 KiB

BIN
media/tutorial/img31.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 84 KiB

BIN
media/tutorial/img32.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 88 KiB

Some files were not shown because too many files have changed in this diff Show More