SAVRN
Search Contact SAVRN

Open-weight model · Text to 3d

UniMate-Weights

by Tarn59 tarn59/UniMate-Weights

UniMate-Weights is an open-weight model for text to 3d from Tarn59, released under MIT License. It has 74M parameters. At 16-bit it needs about 0.2 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 83 downloads a month.

A from-scratch training run of the main UniMate model ("UniMate: One Unified Model to Animate Diverse Skeletons", Mou et al., SIGGRAPH Asia 2026, arXiv 2609.05415) on a single GPU, using the authors' released code and the shipped…

Parameters74M
Context
Weights3.6 GB
Licensemit
AccessOpen weights
Monthly Downloads83

Runs On

What it takes to serve UniMate-Weights (74M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 0.1 GB 0.2 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.1 GB 0.1 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.0 GB 0.0 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 24, 2026.

UniMate-Weights on every accelerator the SAVRN Index prices, at every precision

Model Card

By Tarn59, published under mit, revision 518325e09a55.

UniMate uniml3d_60frames_graph_adaln — independently trained checkpoint (v2)

A from-scratch training run of the main UniMate model ("UniMate: One Unified Model to Animate Diverse Skeletons", Mou et al., SIGGRAPH Asia 2026, arXiv 2609.05415) on a single GPU, using the authors' released code and the shipped configs/uniml3d_60frames_graph_adaln.json config unchanged.

This repository holds two runs. The files at the root are v2, trained on an audited copy of the dataset from which 13 defective clips were removed and 18 repaired (see Why v2). The earlier v1 run on the unaudited data is kept under v1/ for comparison. Use v2.

Read the full model card (1,894 words)

Identity and Version

Repository
tarn59/UniMate-Weights
Publisher
Tarn59
Task
Text to 3d
Modality
Other
Library
pytorch
Parameters
74M parameters
Languages
Not stated by the source
Revision
518325e09a555f059ba0efe6face299e9b853d8c
First published
2026-09-19
Last updated
2026-09-24

Files and Weights

39 files, 3.6 GB in total. The weights are 6 files totalling 3.6 GB in pt, safetensors.

Weights6 files · 3.6 GB
Configuration8 files · 39.8 KB
Documentation4 files · 26.4 KB
Other20 files · 5.7 MB
Repository1 file · 1.9 KB
Every file
FileTypeSizeSHA-256
checkpoints/checkpoint_step_120000.ptWeights1.2 GB ee58a7c08b59
model.safetensorsWeights296.4 MB c871afe2991a
model_ema.safetensorsWeights296.4 MB 716b8b359216
v1/checkpoints/checkpoint_step_120000.ptWeights1.2 GB 116fbc42f243
v1/model.safetensorsWeights296.4 MB d0a96d7355eb
v1/model_ema.safetensorsWeights296.4 MB ac7027bde593
audit/patch_log.jsonConfiguration3.9 KB
audit/patch_poison.pyConfiguration5.3 KB
audit/poison_tiers.jsonConfiguration12.9 KB
audit/score_clips.pyConfiguration6.7 KB
config.jsonConfiguration3.2 KB
train_config_as_launched.jsonConfiguration2.3 KB
v1/config.jsonConfiguration3.2 KB
v1/train_config_as_launched.jsonConfiguration2.3 KB
LICENSEDocumentation1.1 KB
README.mdDocumentation14.1 KB
audit/POISON_REPORT.mdDocumentation10.6 KB
v1/README.mdDocumentation628 B
audit/applied_skip_list_objaverse.txtOther1.5 KB
audit/clip_losses.csvOther2.3 MB
audit/filtered_clips_objaverse.txtOther3.5 KB
audit/filtered_clips_truebones.txtOther426 B
audit/gacha07_preview.pngOther156.5 KB 5c566cf19735
audit/santa_running_preview.pngOther129.6 KB 83b751af5bd0
dataset_stats.npyOther1.9 KB 653a9c7c2ec7
examples/mixamo_leaps_up_and_hangs.mp4Other99.6 KB
examples/objaverse_669a7e6e_scuttles_sideways.mp4Other84.9 KB
examples/objaverse_7f020c8e_dances_in_place.mp4Other103.1 KB ae1930f2e316
examples/truebones_eagle_sits_spreads_wings.mp4Other95.2 KB
logs/events.out.tfevents.1790095024.stuporComputer.510986.0Other921.8 KB 350e93d2374f
training_curves.pngOther219.2 KB 6e49164fa566
v1/dataset_stats.npyOther1.9 KB d24f7f601329
v1/examples/mixamo_jumps_while_running_arm_extended.mp4Other95.8 KB
v1/examples/objaverse_669a73d2_sways_hips_and_legs.mp4Other97.9 KB
v1/examples/objaverse_eeb9bae9_backflip.mp4Other119.0 KB abbea91a8a32
v1/examples/truebones_crocodile_bounces_then_collapses.mp4Other98.3 KB
v1/logs/events.out.tfevents.1789699593.stuporComputer.464765.0Other921.8 KB c3aba82b795e
v1/training_curves.pngOther221.5 KB b757bc34c3b6
.gitattributesRepository1.9 KB

License and Download

License
mit
Access
Open weights, no gate
Download size
3.6 GB
Download from Tarn59

Released by Tarn59 through its official repository on Hugging Face. Read the license.

Built From

  • Described by arXiv:2609.05415
  • Trained on (disclosed) Linzhan/UniML3D

Memory Requirements

PrecisionWeights in memory
As published3.6 GB
16-bit0.1 GB
8-bit0.1 GB
4-bit0.0 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About UniMate-Weights

How much GPU memory does UniMate-Weights need?

About 0.2 GB at 16-bit and 0 GB at 4-bit: the weights (74M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run UniMate-Weights on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use UniMate-Weights commercially?

Yes. UniMate-Weights is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.