Add model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: diffusers
|
| 3 |
+
license: other
|
| 4 |
+
license_name: research-use
|
| 5 |
+
tags:
|
| 6 |
+
- object-detection
|
| 7 |
+
- bounding-boxes
|
| 8 |
+
- grounding
|
| 9 |
+
- diffusion
|
| 10 |
+
- synthetic-data
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# socknetq
|
| 14 |
+
|
| 15 |
+
Joint image + bounding-box diffusion model with a LoRA adapter, a confidence stream, and Nash-MTL multi-task loss weighting.
|
| 16 |
+
|
| 17 |
+
Released by [Breakpoint AI](https://huggingface.co/BreakpointAI) as part of open-sourcing the company's research
|
| 18 |
+
artifacts.
|
| 19 |
+
|
| 20 |
+
| | |
|
| 21 |
+
|---|---|
|
| 22 |
+
| Training data | [`BreakpointAI/breakpoint-grounding-55m`](https://huggingface.co/datasets/BreakpointAI/breakpoint-grounding-55m) |
|
| 23 |
+
| Checkpoint step | 3,120,000 |
|
| 24 |
+
| Training run | [W&B](https://wandb.ai/diffusionexp/train_socknetq/runs/7pqundo1) |
|
| 25 |
+
|
| 26 |
+
## Contents
|
| 27 |
+
|
| 28 |
+
| Path | Size | What it is |
|
| 29 |
+
|---|---|---|
|
| 30 |
+
| `pytorch_model/` | 67.6 GB | |
|
| 31 |
+
| `boxnet/` | 20.1 GB | Joint image + bounding-box diffusion backbone |
|
| 32 |
+
| `pytorch_lora_weights.safetensors` | 1.0 GB | LoRA adapter weights |
|
| 33 |
+
| `conf_weighter.pth` | 2.0 KB | Confidence-stream loss-weighting head |
|
| 34 |
+
| `weighter.pth` | 1.9 KB | Loss-weighting head |
|
| 35 |
+
| `nash_mtl_weights_conf_ema.safetensors` | 120 B | Nash-MTL task-weighting coefficients |
|
| 36 |
+
| `nash_mtl_weights_ema.safetensors` | 112 B | Nash-MTL task-weighting coefficients |
|
| 37 |
+
| `global_conf_step.txt` | 7 B | Training step counter |
|
| 38 |
+
| `global_nonull_step.txt` | 7 B | Training step counter |
|
| 39 |
+
| `weighter_steps.txt` | 4 B | Loss-weighter step counter |
|
| 40 |
+
|
| 41 |
+
**Inference weights only.** Optimizer, LR scheduler, RNG and dataloader state were not
|
| 42 |
+
uploaded, so this checkpoint cannot be used to resume training.
|
| 43 |
+
|
| 44 |
+
## Citation
|
| 45 |
+
|
| 46 |
+
```bibtex
|
| 47 |
+
@misc{breakpoint_socknetq,
|
| 48 |
+
title = {socknetq},
|
| 49 |
+
author = {Wang, Franklin and Zhong, Desmond and Murdoch, Jamie},
|
| 50 |
+
year = {2026},
|
| 51 |
+
url = {https://huggingface.co/BreakpointAI/socknetq}
|
| 52 |
+
}
|
| 53 |
+
```
|