Model reference · open weights
timesformer_GP_scroll1 is an open-weight embedding model from scrollprize. timesformer_GP_scroll1 (FP32) weighs 76 MB; the smallest configuration that runs it is RTX 3060 12 GB.
For self-hosting, compare this build's requirements below with the GPU you plan to rent.
Summary of the scrollprize/timesformer_GP_scroll1 model card, 2026-10-01
What it is
| Released by | scrollprize |
|---|---|
| Released | 2024-12-12 |
| Parameters | 38M |
| VRAM | 76 MB for the weights |
What it runs on
| Card | Runs | Memory |
|---|---|---|
| RTX 3060 12 GB | fits | 11.6 GB |
| RTX 4060 Ti 16 GB | fits | 15.4 GB |
| RTX 3090 24 GB | fits | 23.4 GB |
| RTX 4090 24 GB | fits | 23.4 GB |
| RTX 5090 32 GB | fits | 31.0 GB |
| L40S 48 GB | fits | 44.0 GB |
| A100 80 GB | fits | 78.2 GB |
| H100 80 GB | fits | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | fits | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | fits | 107 GB |
| H200 141 GB | fits | 138 GB |
| B200 180 GB | fits | 176 GB |
From the model card
The grandprize winning model of the Vesuvius Challenge of 2023.
The grandprize winning model of the Vesuvius Challenge of 2023. The model features a small TimeSformer architecture trained on image segmentation task to detect ink in 3d images. This model takes as input the 3d image and outputs a 2d map of ink detections, roughly 1/16 the size of the input.
Make sure to have the dependencies installed, namely transformers and Timesformer package
pip install -U transformers timesformer-pytorch
Next you can run the model as follows:
from transformers import AutoModel
model = AutoModel.from_pretrained("YoussefMoNader/timesformer_GP_scroll1", trust_remote_code=True)
the model expects a (B,1,26,64,64) tensor
The model was trained on 4xH100 for 8 hours. This model was trained for 12 epochs on total, a single epoch takes around 45 mins using the old script train_timesformer_og.py
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.