fix(annotate): transcode subclips to H.264 instead of stream-copy

Modern LeRobot datasets store videos in AV1, which vllm's libav build
cannot decode (the video processor returns 0 frames and downstream
chokes with ZeroDivisionError). Re-encode each per-episode subclip
with libx264 (preset ultrafast, crf 23) so the resulting mp4 is
universally decodable. Strip audio with -an for a smaller payload.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

pyproject.toml

@@ -302,7 +302,7 @@ lerobot-annotate="lerobot.scripts.lerobot_annotate:main"
 
 # ---------------- Tool Configurations ----------------
 [tool.setuptools.package-data]
-lerobot = ["envs/*.json"]
+lerobot = ["envs/*.json", "annotations/steerable_pipeline/prompts/*.txt"]
 
 [tool.setuptools.packages.find]
 where = ["src"]

src/lerobot/annotations/steerable_pipeline/config.py

@@ -18,6 +18,7 @@ from __future__ import annotations
 
 from dataclasses import dataclass, field
 from pathlib import Path
+from typing import Any
 
 
 @dataclass
@@ -107,10 +108,17 @@ class VlmConfig:
     this command (if present) is substituted per-replica."""
     parallel_servers: int = 1
     """When >1, spawn this many independent inference servers (each pinned
-    to one GPU via ``CUDA_VISIBLE_DEVICES`` and listening on
+    to a GPU via ``CUDA_VISIBLE_DEVICES`` and listening on
     ``serve_port + i``) and round-robin client requests across them.
     Useful when DP/TP NCCL setup is broken on the node — single-GPU
-    replicas don't need cross-GPU communication."""
+    replicas don't need cross-GPU communication. When
+    ``parallel_servers > num_gpus``, replicas are round-robin-assigned
+    to GPUs (e.g. 4 replicas on 2 GPUs → 0,1,0,1)."""
+    num_gpus: int = 0
+    """How many physical GPUs are available for round-robin replica
+    placement. ``0`` means ``parallel_servers`` (one GPU per replica,
+    backward-compatible default). Set this to ``2`` with
+    ``parallel_servers=4`` to pack 2 replicas per GPU."""
     client_concurrency: int = 16
     """Maximum number of in-flight chat requests the client issues in
     parallel. vllm batches them internally for free, so bumping this
@@ -140,6 +148,12 @@ class VlmConfig:
     camera_key: str | None = None
     """Override the camera stream used for keyframe attachment. ``None`` picks
     the first ``observation.images.*`` key the dataset declares."""
+    chat_template_kwargs: dict[str, Any] | None = None
+    """Forwarded as ``extra_body.chat_template_kwargs`` on every chat call.
+    Use this to pass model-specific template flags such as
+    ``{"enable_thinking": false}`` for Qwen3.5/Qwen3.6 to suppress the
+    reasoning preamble that otherwise eats the entire ``max_new_tokens``
+    budget before any JSON is emitted."""
 
 
 @dataclass

src/lerobot/annotations/steerable_pipeline/frames.py

@@ -225,8 +225,11 @@ def episode_clip_path(
     """Extract the episode's subclip to ``cache_dir/ep_{idx:06d}.mp4``.
 
     Returns ``None`` if the dataset has no video tracks. Skips re-extract
-    when the cached clip already exists. Uses ``ffmpeg`` via subprocess
-    with stream-copy where possible (no re-encode) for speed.
+    when the cached clip already exists. Re-encodes to H.264
+    (libx264) so the resulting mp4 is decodable by every downstream
+    video processor — stream-copy would inherit the source codec
+    (often AV1 in modern LeRobot datasets), which vllm's libav build
+    cannot decode.
     """
     import subprocess  # noqa: PLC0415
 
@@ -253,12 +256,19 @@ def episode_clip_path(
         f"{to_timestamp:.3f}",
         "-i",
         str(src),
-        "-c",
-        "copy",
+        "-c:v",
+        "libx264",
+        "-preset",
+        "ultrafast",
+        "-crf",
+        "23",
+        "-pix_fmt",
+        "yuv420p",
+        "-an",
         str(out_path),
     ]
     try:
-        subprocess.run(cmd, check=True, timeout=120)
+        subprocess.run(cmd, check=True, timeout=300)
     except (subprocess.CalledProcessError, subprocess.TimeoutExpired, FileNotFoundError):
         return None
     return out_path if out_path.exists() and out_path.stat().st_size > 0 else None

src/lerobot/annotations/steerable_pipeline/vlm_client.py

@@ -380,10 +380,13 @@ def _make_openai_client(config: VlmConfig) -> VlmClient:
             "max_tokens": max_tok,
             "temperature": temp,
         }
+        extra_body: dict[str, Any] = {}
         if send_mm_kwargs and mm_kwargs:
-            kwargs["extra_body"] = {
-                "mm_processor_kwargs": {**mm_kwargs, "do_sample_frames": True}
-            }
+            extra_body["mm_processor_kwargs"] = {**mm_kwargs, "do_sample_frames": True}
+        if config.chat_template_kwargs:
+            extra_body["chat_template_kwargs"] = config.chat_template_kwargs
+        if extra_body:
+            kwargs["extra_body"] = extra_body
         with rr_lock:
             chosen = clients[rr_counter["i"] % len(clients)]
             rr_counter["i"] += 1
@@ -453,10 +456,12 @@ def _spawn_parallel_inference_servers(config: VlmConfig) -> list[str]:
         f"--uvicorn-log-level warning"
     )
 
+    num_gpus = config.num_gpus if config.num_gpus > 0 else n
     for i in range(n):
         port = config.serve_port + i
+        gpu = i % num_gpus
         env = _os.environ.copy()
-        env["CUDA_VISIBLE_DEVICES"] = str(i)
+        env["CUDA_VISIBLE_DEVICES"] = str(gpu)
         cmd = base_cmd
         if "{port}" in cmd:
             cmd = cmd.replace("{port}", str(port))
@@ -464,7 +469,7 @@ def _spawn_parallel_inference_servers(config: VlmConfig) -> list[str]:
             cmd = f"{cmd} --port {port}"
         api_base = f"http://localhost:{port}/v1"
         api_bases.append(api_base)
-        print(f"[server-{i}] launching on GPU {i} port {port}: {cmd}", flush=True)
+        print(f"[server-{i}] launching on GPU {gpu} port {port}: {cmd}", flush=True)
         proc = subprocess.Popen(
             shlex.split(cmd),
             stdout=subprocess.PIPE,