Magma: A Foundation Model for Multimodal AI Agents
Paper • 2502.13130 • Published • 58
dataset_name stringclasses 1
value | video_name stringlengths 41 48 | start_time float32 0 308 | end_time float32 0 310 | task_string stringlengths 9 159 | frame_index int32 0 5.7k | height int32 512 1.02k | width int32 512 1.36k | trace unknown | trace_visibility unknown |
|---|---|---|---|---|---|---|---|---|---|
ego4d | 16a8e06b-aff1-445e-b942-12de75f35519_1620_1920 | 127.317009 | 133.609711 | the camera wearer spreads sand with a trowel | 0 | 512 | 682 | [
128,
4,
149,
147,
128,
0,
0,
0,
0,
0,
0,
140,
21,
110,
117,
109,
112,
121,
46,
99,
111,
114,
101,
46,
109,
117,
108,
116,
105,
97,
114,
114,
97,
121,
148,
140,
12,
95,
114,
101,
99,
111,
110,
115,
116,
114,
117,
99,
116,... | [
128,
4,
149,
147,
16,
0,
0,
0,
0,
0,
0,
140,
21,
110,
117,
109,
112,
121,
46,
99,
111,
114,
101,
46,
109,
117,
108,
116,
105,
97,
114,
114,
97,
121,
148,
140,
12,
95,
114,
101,
99,
111,
110,
115,
116,
114,
117,
99,
116,
... |
ego4d | 538c2f0a-f89e-47b5-b012-047a1b98de27_1343_1649 | 96.546608 | 119.401939 | the camera wearer is chopping chili | 528 | 512 | 682 | [
128,
4,
149,
147,
128,
0,
0,
0,
0,
0,
0,
140,
21,
110,
117,
109,
112,
121,
46,
99,
111,
114,
101,
46,
109,
117,
108,
116,
105,
97,
114,
114,
97,
121,
148,
140,
12,
95,
114,
101,
99,
111,
110,
115,
116,
114,
117,
99,
116,... | [
128,
4,
149,
147,
16,
0,
0,
0,
0,
0,
0,
140,
21,
110,
117,
109,
112,
121,
46,
99,
111,
114,
101,
46,
109,
117,
108,
116,
105,
97,
114,
114,
97,
121,
148,
140,
12,
95,
114,
101,
99,
111,
110,
115,
116,
114,
117,
99,
116,
... |
ego4d | 998100d5-7fc9-40fd-b136-54eb3aa7628a_809_1109 | 115.557823 | 144.599487 | the camera wearer mixes food | 400 | 512 | 682 | [
128,
4,
149,
147,
128,
0,
0,
0,
0,
0,
0,
140,
21,
110,
117,
109,
112,
121,
46,
99,
111,
114,
101,
46,
109,
117,
108,
116,
105,
97,
114,
114,
97,
121,
148,
140,
12,
95,
114,
101,
99,
111,
110,
115,
116,
114,
117,
99,
116,... | [
128,
4,
149,
147,
16,
0,
0,
0,
0,
0,
0,
140,
21,
110,
117,
109,
112,
121,
46,
99,
111,
114,
101,
46,
109,
117,
108,
116,
105,
97,
114,
114,
97,
121,
148,
140,
12,
95,
114,
101,
99,
111,
110,
115,
116,
114,
117,
99,
116,
... |
ego4d | f6277269-1c87-439c-b5be-d4a02343018a_269_570 | 279.486053 | 289.949524 | the camera wearer washes the body of the car with the foam in his right hand | 288 | 512 | 677 | "gASVk4AAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) | "gASVkxAAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) |
ego4d | 392b632b-40b1-4fb0-92cf-dc17044472ca_540_840 | 258.227661 | 262.52356 | the camera wearer unfolds the textile | 0 | 512 | 677 | "gASVk4AAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) | "gASVkxAAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) |
ego4d | 6e4baf27-2090-4a90-816b-a35a0aff0e56_2429_2729 | 70.641319 | 79.913628 | the camera wearer rolls the dough on the mat | 64 | 512 | 677 | "gASVk4AAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) | "gASVkxAAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) |
ego4d | 4d2d3ce2-0d55-4a5e-9af0-5d4a37e05f72_2430_2730 | 17.175461 | 43.584991 | the camera wearer stops painting | 512 | 512 | 682 | "gASVk4AAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) | "gASVkxAAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) |
ego4d | 4b63760e-3016-4a13-ad93-c9487a433a4c_1350_1602 | 160.958038 | 165.428726 | the camera wearer paints on a paper | 32 | 512 | 682 | "gASVk4AAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) | "gASVkxAAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) |
ego4d | 12946f28-5f4a-4ba5-8296-c08cbeeca501_0_300 | 281.67804 | 283.987823 | the camera wearer puts the knife on the sink | 16 | 512 | 682 | "gASVk4AAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) | "gASVkxAAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) |
ego4d | 38ead8a1-4d57-4bfb-b7dc-a9785b22f93a_0_300 | 45.827831 | 47.571102 | the camera wearer pours liquid in a toilet | 0 | 512 | 677 | "gASVk4AAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) | "gASVkxAAAAAAAACMFW51bXB5LmNvcmUubXVsdGlhcnJheZSMDF9yZWNvbnN0cnVjdJSTlIwFbnVtcHmUjAduZGFycmF5lJOUSwC(...TRUNCATED) |
End of preview. Expand in Data Studio
Magma: A Foundation Model for Multimodal AI Agents
Jianwei Yang*1† Reuben Tan1† Qianhui Wu1† Ruijie Zheng2‡ Baolin Peng1‡ Yongyuan Liang2‡
Yu Gu1 Mu Cai3 Seonghyeon Ye4 Joel Jang5 Yuquan Deng5 Lars Liden1 Jianfeng Gao1▽
1 Microsoft Research; 2 University of Maryland; 3 University of Wisconsin-Madison
4 KAIST; 5 University of Washington
* Project lead † First authors ‡ Second authors ▽ Leadership
[arXiv Paper] [Project Page] [Hugging Face Paper] [Github Repo] [Video]
Introduction
This dataset contains the robotic manipulation data used in Magma pretraining. For fair comparison, we followed OpenVLA to use the data mix "siglip-224px+mx-oxe-magic-soup".
The dataset is organized by following source datasets, with each source containing one or more arrow files:
| Folder | Number of Shards |
|---|---|
| ego4d | 15 |
| sthv2 | 6 |
| instruct_video | 14 |
Features
In addition to the default features, we extracted the visual traces of future 16 frames for each frame. The dataset contains the following fields:
dataset_name: Original source dataset namevideo_name: video nametask_string: Description of the task- 'start_time': starting time stamp for the video segment
- 'end_time': ending time stamp for the video segment
frame_index: starting index of the frame in the video segmentheight: resized image height for visual trace extraction- 'width': resized image width for visual trace extraction
trace: Robot trajectory trace (serialized numpy array)trace_visibility: Visibility mask for the trace (serialized numpy array)
Dataset Loading
Full Dataset Load
from datasets import load_dataset
dataset = load_dataset("MagmaAI/Magma-Video-ToM", streaming=True, split="train")
Individual Dataset Load
or specify a dataset by:
from datasets import load_dataset
dataset = load_dataset("MagmaAI/Magma-Video-ToM", data_dir="sthv2", streaming=True, split="train")
Sample Decoding
# Helper function to deserialize binary fields
def deserialize_array(bytes_data):
return pickle.loads(bytes_data)
# Helper function to convert binary image data to PIL Image
def bytes_to_image(image_bytes):
return Image.open(io.BytesIO(image_bytes))
for i, example in enumerate(dataset):
# decode trace: 1 x 16 x 256 x 2
trace = deserialize_array(example['trace'])
# decode trace visibility: 1 x 16 x 256 x 1
trace_visibility = deserialize_array(example['trace_visibility'])
NOTE: the temporal length of traces for video data is 16 as we excluded the starting frame. For all robotics data, it is 17 as we did not exclude the starting frame.
- Downloads last month
- 79