Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos
Paper • 2206.11795 • Published
Datasets:
video_id stringclasses 121
values | frame_idx int32 0 1.5k | action stringlengths 49 115 | image unknown |
|---|---|---|---|
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 0 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | [
255,
216,
255,
224,
0,
16,
74,
70,
73,
70,
0,
1,
2,
0,
0,
1,
0,
1,
0,
0,
255,
254,
0,
16,
76,
97,
118,
99,
54,
49,
46,
49,
57,
46,
49,
48,
49,
0,
255,
219,
0,
67,
0,
8,
4,
4,
4,
4,
4,
5,
5,
5,
5,
5,
5,
6,
... |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 1 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | [
255,
216,
255,
224,
0,
16,
74,
70,
73,
70,
0,
1,
2,
0,
0,
1,
0,
1,
0,
0,
255,
254,
0,
16,
76,
97,
118,
99,
54,
49,
46,
49,
57,
46,
49,
48,
49,
0,
255,
219,
0,
67,
0,
8,
4,
4,
4,
4,
4,
5,
5,
5,
5,
5,
5,
6,
... |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 2 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | "/9j/4AAQSkZJRgABAgAAAQABAAD//gAQTGF2YzYxLjE5LjEwMQD/2wBDAAgEBAQEBAUFBQUFBQYGBgYGBgYGBgYGBgYHBwcICAg(...TRUNCATED) |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 3 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | "/9j/4AAQSkZJRgABAgAAAQABAAD//gAQTGF2YzYxLjE5LjEwMQD/2wBDAAgEBAQEBAUFBQUFBQYGBgYGBgYGBgYGBgYHBwcICAg(...TRUNCATED) |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 4 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | "/9j/4AAQSkZJRgABAgAAAQABAAD//gAQTGF2YzYxLjE5LjEwMQD/2wBDAAgEBAQEBAUFBQUFBQYGBgYGBgYGBgYGBgYHBwcICAg(...TRUNCATED) |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 5 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | "/9j/4AAQSkZJRgABAgAAAQABAAD//gAQTGF2YzYxLjE5LjEwMQD/2wBDAAgEBAQEBAUFBQUFBQYGBgYGBgYGBgYGBgYHBwcICAg(...TRUNCATED) |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 6 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | "/9j/4AAQSkZJRgABAgAAAQABAAD//gAQTGF2YzYxLjE5LjEwMQD/2wBDAAgEBAQEBAUFBQUFBQYGBgYGBgYGBgYGBgYHBwcICAg(...TRUNCATED) |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 7 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | "/9j/4AAQSkZJRgABAgAAAQABAAD//gAQTGF2YzYxLjE5LjEwMQD/2wBDAAgEBAQEBAUFBQUFBQYGBgYGBgYGBgYGBgYHBwcICAg(...TRUNCATED) |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 8 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | "/9j/4AAQSkZJRgABAgAAAQABAAD//gAQTGF2YzYxLjE5LjEwMQD/2wBDAAgEBAQEBAUFBQUFBQYGBgYGBgYGBgYGBgYHBwcICAg(...TRUNCATED) |
bumpy-pumpkin-dunker-5c7c23a408bf-20220104-193828 | 9 | <|action_start|> 0 0 0 ; LMB ; LMB ; LMB ; LMB <|action_end|> | "/9j/4AAQSkZJRgABAgAAAQABAAD//gAQTGF2YzYxLjE5LjEwMQD/2wBDAAgEBAQEBAUFBQUFBQYGBgYGBgYGBgYGBgYHBwcICAg(...TRUNCATED) |
End of preview. Expand in Data Studio
Minecraft VLA Stage 1: Action Pretraining Data
Vision-Language-Action training data for Minecraft, processed from OpenAI's VPT contractor dataset.
Dataset Description
This dataset contains frame-action pairs from Minecraft gameplay, designed for training VLA models following the Lumine methodology.
Source
- Original: OpenAI VPT Contractor Data (7.x subset)
- Videos:
17,886 videos (330 hours of early-game gameplay) - Task: "Play Minecraft" with focus on first 30 minutes of new worlds
Format
Each sample contains:
| Field | Type | Description |
|---|---|---|
image |
bytes | 640x360 JPEG frame |
video_id |
string | Source video identifier |
frame_idx |
int | Frame number at 5Hz |
action |
string | Lumine-format action string |
Action Format
<|action_start|> mouse_x mouse_y scroll ; K1 ; K2 ; K3 ; K4 <|action_end|>
mouse_x,mouse_y: Mouse delta (-1000 to 1000)scroll: Hotbar scroll (always 0 - VPT uses number keys)K1toK4: Key combinations per 50ms chunk
Example:
<|action_start|> 45 -12 0 ; W ; W Space ; W LMB ; W LMB <|action_end|>
Processing Details
- Frame rate: 5 FPS (downsampled from VPT's 20 FPS)
- Action chunks: 4 per frame (each 50ms = 200ms total)
- Filtering: Idle frames removed, loading screens filtered
Usage
from datasets import load_dataset
# Streaming (recommended - no download required)
ds = load_dataset("TESS-Computer/minecraft-vla-stage1", split="train", streaming=True)
for sample in ds:
image = sample["image"] # PIL Image or bytes
action = sample["action"]
# Process...
Training Pipeline
This is Stage 1 of a 3-stage training pipeline:
- Stage 1 (this dataset): Action pretraining - learn observation→action mapping
- Stage 2: Instruction following - add task instructions from JARVIS-VLA
- Stage 3: Reasoning - add chain-of-thought before complex actions
Citation
If you use this dataset, please cite:
- OpenAI VPT - Original contractor data
- JARVIS-VLA - Instruction annotations
- Lumine - Training methodology
License
MIT License. Original VPT data is released under MIT by OpenAI.
- Downloads last month
- 34