admin/lerobot
1
0
Fork 0
mirror of https://github.com/huggingface/lerobot.git synced 2026-05-26 05:59:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

small fix

Browse Source
This commit is contained in:
Pepijn
2025-11-18 13:47:05 +01:00
parent cf0f878dbb
commit 3c9149e909
1 changed files with 9 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/dataset_annotation/subtask_annotation.py
+9 -9
View File
@@ -43,7 +43,7 @@ Usage:
pip install transformers torch qwen-vl-utils accelerate pip install transformers torch qwen-vl-utils accelerate
# Annotate and push to hub: # Annotate and push to hub:
python subtask_annotation_local.py \\ python subtask_annotation.py \\
--repo-id pepijn223/mydataset \\ --repo-id pepijn223/mydataset \\
--subtasks "reach,grasp,lift,place" \\ --subtasks "reach,grasp,lift,place" \\
--video-key observation.images.base \\ --video-key observation.images.base \\
@@ -65,7 +65,7 @@ from rich.console import Console
from rich.panel import Panel from rich.panel import Panel
from rich.progress import Progress, SpinnerColumn, TextColumn from rich.progress import Progress, SpinnerColumn, TextColumn
from rich.tree import Tree from rich.tree import Tree
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor from transformers import Qwen3VLMoeForConditionalGeneration, AutoProcessor
from lerobot.datasets.lerobot_dataset import LeRobotDataset from lerobot.datasets.lerobot_dataset import LeRobotDataset
@@ -158,7 +158,7 @@ class VideoAnnotator:
self.console.print(f"[cyan]Loading model: {model_name}...[/cyan]") self.console.print(f"[cyan]Loading model: {model_name}...[/cyan]")
# Load model and processor # Load model and processor
self.model = Qwen2VLForConditionalGeneration.from_pretrained( self.model = Qwen3VLMoeForConditionalGeneration.from_pretrained(
model_name, model_name,
torch_dtype=torch_dtype, torch_dtype=torch_dtype,
device_map=device, device_map=device,
@@ -743,15 +743,15 @@ def main():
epilog=""" epilog="""
Examples: Examples:
# List available cameras: # List available cameras:
python subtask_annotation_local.py --repo-id pepijn223/mydataset --subtasks "reach,grasp" --max-episodes 0 python subtask_annotation.py --repo-id pepijn223/mydataset --subtasks "reach,grasp" --max-episodes 0
# Annotate with specific camera: # Annotate with specific camera:
python subtask_annotation_local.py --repo-id pepijn223/mydataset --subtasks "reach,grasp" --video-key observation.images.top --push-to-hub python subtask_annotation.py --repo-id pepijn223/mydataset --subtasks "reach,grasp" --video-key observation.images.top --push-to-hub
# Use smaller model (7B instead of 30B): # Use custom model:
python subtask_annotation_local.py --repo-id pepijn223/mydataset --subtasks "reach,grasp" --video-key observation.images.top --model Qwen/Qwen2-VL-7B-Instruct --push-to-hub python subtask_annotation.py --repo-id pepijn223/mydataset --subtasks "reach,grasp" --video-key observation.images.top --model Qwen/Qwen3-VL-30B-A3B-Instruct --push-to-hub
Note: The 7B model requires ~16GB VRAM. Use 2B model (~8GB VRAM) if needed. Note: The 30B model requires ~60GB VRAM. Make sure you have sufficient GPU memory.
""" """
) )
parser.add_argument( parser.add_argument(
@@ -783,7 +783,7 @@ Note: The 7B model requires ~16GB VRAM. Use 2B model (~8GB VRAM) if needed.
"--model", "--model",
type=str, type=str,
default="Qwen/Qwen3-VL-30B-A3B-Instruct", default="Qwen/Qwen3-VL-30B-A3B-Instruct",
help="Model to use (default: Qwen/Qwen3-VL-30B-A3B-Instruct). Other options: Qwen/Qwen2-VL-2B-Instruct, Qwen/Qwen2-VL-7B-Instruct", help="Qwen3-VL model to use (default: Qwen/Qwen3-VL-30B-A3B-Instruct)",
) )
parser.add_argument( parser.add_argument(
"--skip-existing", "--skip-existing",