mirror of
https://github.com/huggingface/lerobot.git
synced 2026-06-04 04:41:24 +00:00
eba3ab3741
Bugs
* validator: don't re-raise on unknown style. The second column_for_style
lookup (used to route persistent vs event) now sits in try/except so an
unknown style is recorded by _check_column_routing and skipped instead
of crashing the whole validation pass.
* general_vqa._target_cameras: when restrict_to_default_camera is set but
the configured camera_key isn't one the provider exposes, warn and fall
back to all cameras instead of returning a phantom key that KeyErrors
deep in frame decode.
* interjections: clamp interjection timestamps to frame_timestamps[0]
rather than a hardcoded 0.0 (datasets can start at non-zero t).
Docs / code drift
* annotation_pipeline.mdx: drop the phantom 'vocabulary discovery / phase
0 / --vocabulary.* / canonical_vocabulary.json' section (none of it
exists); describe the real describe->segment + coverage-stitch flow.
Soften the src/lerobot/tools/ + TOOL_REGISTRY reference to 'not part of
this PR' (matches tools.mdx, which already marks the runtime layer as
not-yet-implemented). Fix the --push_to_hub/--new_repo_id wording. Note
the default is now a single h200. Add a 'Contributing new modules'
section inviting module / prompt / quality contributions.
* executor docstring: six phases, no phantom phase 0.
run_hf_job.py
* add the Apache 2.0 license header (was flagged repeatedly).
* default to a single GPU: flavor=h200, parallel_servers=1, num_gpus=1
(scale to h200x4 noted in the docstring).
* pin the install to @main instead of the feature branch (won't break
after merge).
Naming / cleanup
* rename dest_repo_id -> new_repo_id across config / script / example /
test to match the LeRobot dataset edit tools.
* rename prompt templates module_N_*.txt -> descriptive (plan_*,
interjections_*, vqa.txt) and update every load_prompt() call.
* remove dead _messages_to_prompt (used only by the removed in-process
backends).
* declare _warned_decode_fail (frames) and _warned_no_camera (vqa) as
real init=False dataclass fields instead of getattr monkey-patches.
* scope bandit B607 to the two ffmpeg subprocess.run sites via
'# nosec B607' and drop it from the global skip list.
Tests
* fix stale canned-VLM markers ('ONE realistic interruption' ->
'compact interjection', 'Update the memory' -> 'compressed semantic
memory') and drop the dead 'concise hierarchical PLAN' plan responders
(plan generation is deterministic now) in run_e2e_smoke,
test_pipeline_recipe_render, test_modules.
* run_e2e_smoke now asserts interjection + speech rows are produced so a
stale marker can't silently pass again.
* drop remaining 'PR 1' / 'PR 2' references from test comments / names.
Verified: tests/annotations + tests/datasets/test_language +
tests/scripts/test_lerobot_annotate (31 passed); make-style E2E smoke
(interjections=1 speech_atoms=2); pre-commit (ruff, mypy, bandit,
prettier) clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
136 lines
6.2 KiB
Python
136 lines
6.2 KiB
Python
#!/usr/bin/env python
|
|
|
|
# Copyright 2026 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.
|
|
"""Launch ``lerobot-annotate`` on a Hugging Face job (vllm + Qwen3.6-27B VLM).
|
|
|
|
Spawns one single-GPU ``h200`` job that:
|
|
|
|
1. installs ``lerobot`` plus the annotation extras,
|
|
2. boots one vllm server with Qwen3.6-27B (dense VLM),
|
|
3. runs the plan / interjections / vqa modules across the dataset
|
|
in free-form mode (each episode generates its own subtasks +
|
|
memory),
|
|
4. uploads the annotated dataset to ``--new_repo_id`` (when set)
|
|
or back to ``--repo_id``.
|
|
|
|
Usage:
|
|
|
|
HF_TOKEN=hf_... uv run python examples/annotations/run_hf_job.py
|
|
|
|
Adjust ``CMD`` (dataset, model, hub repo) and ``flavor`` below for your
|
|
run. For larger datasets, scale to ``h200x4`` and raise
|
|
``--vlm.parallel_servers`` / ``--vlm.num_gpus`` to match.
|
|
"""
|
|
|
|
import os
|
|
|
|
from huggingface_hub import get_token, run_job
|
|
|
|
token = os.environ.get("HF_TOKEN") or get_token()
|
|
if not token:
|
|
raise RuntimeError("No HF token. Run `huggingface-cli login` or `export HF_TOKEN=hf_...`")
|
|
|
|
CMD = (
|
|
"apt-get update -qq && apt-get install -y -qq git ffmpeg && "
|
|
"pip install --no-deps "
|
|
"'lerobot @ git+https://github.com/huggingface/lerobot.git@main' && "
|
|
"pip install --upgrade-strategy only-if-needed "
|
|
"datasets pyarrow av jsonlines draccus gymnasium torchcodec mergedeep pyyaml-include toml typing-inspect "
|
|
"openai && "
|
|
"export VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0 && "
|
|
"export VLLM_VIDEO_BACKEND=pyav && "
|
|
"lerobot-annotate "
|
|
"--repo_id=pepijn223/robocasa_pretrain_human300_v4 "
|
|
"--new_repo_id=pepijn223/robocasa_pretrain_human300_v4_annotated5 "
|
|
"--push_to_hub=true "
|
|
"--vlm.backend=openai "
|
|
"--vlm.model_id=Qwen/Qwen3.6-27B "
|
|
"--vlm.parallel_servers=1 "
|
|
"--vlm.num_gpus=1 "
|
|
'--vlm.serve_command="vllm serve Qwen/Qwen3.6-27B '
|
|
"--tensor-parallel-size 1 --max-model-len 32768 "
|
|
'--gpu-memory-utilization 0.8 --uvicorn-log-level warning --port {port}" '
|
|
"--vlm.serve_ready_timeout_s=1800 "
|
|
"--vlm.client_concurrency=128 "
|
|
"--vlm.max_new_tokens=512 "
|
|
"--vlm.temperature=0.7 "
|
|
"--executor.episode_parallelism=16 "
|
|
"--vlm.chat_template_kwargs='{\"enable_thinking\": false}' "
|
|
"--vlm.camera_key=observation.images.robot0_agentview_right "
|
|
# Phase 1 — plan module (subtasks + plan + memory).
|
|
# Embed decoded frames directly (use_video_url=false) rather than
|
|
# handing the server a file:// clip. The embedded path is more
|
|
# reliable: if clip extraction ever fails, the video_url path would
|
|
# silently send NO video and the VLM would hallucinate subtasks from
|
|
# the task text alone.
|
|
#
|
|
# CONTEXT BUDGET: with embedded frames, each frame is ~250-320 vision
|
|
# tokens. The model's context is 32768 (see --max-model-len). 32
|
|
# frames sampled uniformly across the episode (~8-10k tokens) fits
|
|
# comfortably alongside the prompt and the describe pass.
|
|
# Do NOT raise max_video_frames toward 128 with embedded frames — that
|
|
# is ~33-39k tokens and overflows the context (BadRequestError 400,
|
|
# "Input length exceeds maximum context length").
|
|
"--plan.use_video_url=false "
|
|
"--plan.frames_per_second=1.0 "
|
|
"--plan.max_video_frames=32 "
|
|
# Constant 1 fps density via windowing: episodes longer than 32s are
|
|
# split into 32-second windows (each 32 frames @ 1 fps, fits context),
|
|
# so long episodes get MORE subtasks instead of a sparser whole-episode
|
|
# view. describe->segment runs per window; spans are merged +
|
|
# stitched to a contiguous whole-episode cover. 0 disables.
|
|
"--plan.subtask_window_seconds=32 "
|
|
# IMPORTANT for RoboCasa: the dataset's task string ("Navigate to the
|
|
# stove", "Pick the mug...") is authoritative and is what eval uses.
|
|
# ``derive_task_from_video=off`` keeps that canonical task driving
|
|
# subtask generation. Do NOT use ``always`` here — it throws the real
|
|
# task away, asks the VLM "what is this video about?" with no hint,
|
|
# and the hallucinated task then poisons every subtask + plan row.
|
|
"--plan.derive_task_from_video=off "
|
|
# NO task augmentation for RoboCasa: eval conditions on the exact task
|
|
# strings, so synthetic rephrasings are unused at best and (when they
|
|
# drift, e.g. "wander around the kitchen") harmful. 0 rephrasings +
|
|
# axes disabled = the policy only ever sees the canonical task.
|
|
"--plan.n_task_rephrasings=0 "
|
|
# action_records OFF: the structured {verb,object,arm,grasp,dest}
|
|
# schema is a manipulation schema; RoboCasa navigation / atomic tasks
|
|
# don't fit it and the VLM hallucinates. When on, records are purely
|
|
# additive (emitted as style="action_record" rows) and never touch
|
|
# the subtask text — useful only for long composite manipulation
|
|
# tasks. Leave off for RoboCasa atomic / navigation.
|
|
# Keep subtask decomposition tight for atomic tasks:
|
|
"--plan.plan_max_steps=10 "
|
|
# Only annotate subtasks + memory — skip the numbered "plan" rows
|
|
# (and their per-boundary VLM call). Flip to true to re-enable plan.
|
|
"--plan.emit_plan=false "
|
|
# NOTE: the grounding pass (describe -> segment, +1 VLM call/episode)
|
|
# is ON BY DEFAULT. Pass --plan.subtask_describe_first=false to disable
|
|
# on datasets you've verified are easy and want fewer calls.
|
|
# Phase 2 — interjections + speech.
|
|
"--interjections.max_interjections_per_episode=6 "
|
|
# Phase 4 — general VQA: DISABLED for this run.
|
|
"--vqa.enabled=false"
|
|
)
|
|
|
|
job = run_job(
|
|
image="vllm/vllm-openai:latest",
|
|
command=["bash", "-c", CMD],
|
|
flavor="h200",
|
|
secrets={"HF_TOKEN": token},
|
|
timeout="2h",
|
|
)
|
|
print(f"Job URL: {job.url}")
|
|
print(f"Job ID: {job.id}")
|