Link public build pipeline + verifier repo, add CI badge
Browse files
README.md
CHANGED
|
@@ -46,6 +46,8 @@ model-index:
|
|
| 46 |
|
| 47 |
<p align="center">
|
| 48 |
<img src="https://huggingface.co/manifesta/adaption_verified_math_code_instruct/resolve/main/assets/banner.png" width="100%" alt="Show Your Work: a verified math and code adapter for Gemma 4 31B">
|
|
|
|
|
|
|
| 49 |
</p>
|
| 50 |
|
| 51 |
[](https://manifesta.adaptionlabs.app/)
|
|
@@ -348,6 +350,27 @@ That +2.5% is the smallest lift of the four datasets built in this project, and
|
|
| 348 |
- [`manifesta/scientific-chart-qa-lora-qwen3.5-9b`](https://huggingface.co/manifesta/scientific-chart-qa-lora-qwen3.5-9b), the first chart build on a 9B base. 51 vs 49, 21 optimizer steps, LoRA on only 8 of 32 decoder layers and again zero vision tensors. Its declared base `togethercomputer/Qwen3.5-9B` also 404s.
|
| 349 |
- [`manifesta/brandvoice-marketing-model`](https://huggingface.co/manifesta/brandvoice-marketing-model), the Part 1 marketing adapter that did work, 56% win rate on Llama 3.3 70B. Style transfer on a base with a weak default style, which is the case where a small adapter has room to move.
|
| 350 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 351 |
## License
|
| 352 |
|
| 353 |
- **Adapter weights:** Apache-2.0.
|
|
|
|
| 46 |
|
| 47 |
<p align="center">
|
| 48 |
<img src="https://huggingface.co/manifesta/adaption_verified_math_code_instruct/resolve/main/assets/banner.png" width="100%" alt="Show Your Work: a verified math and code adapter for Gemma 4 31B">
|
| 49 |
+
<a href="https://github.com/A1VARA5/verified-math-code-17k"><img alt="GitHub" src="https://img.shields.io/badge/GitHub-pipeline%20%2B%20verifier-181717?logo=github"></a>
|
| 50 |
+
<a href="https://github.com/A1VARA5/verified-math-code-17k/actions/workflows/verify.yml"><img alt="verify" src="https://github.com/A1VARA5/verified-math-code-17k/actions/workflows/verify.yml/badge.svg"></a>
|
| 51 |
</p>
|
| 52 |
|
| 53 |
[](https://manifesta.adaptionlabs.app/)
|
|
|
|
| 350 |
- [`manifesta/scientific-chart-qa-lora-qwen3.5-9b`](https://huggingface.co/manifesta/scientific-chart-qa-lora-qwen3.5-9b), the first chart build on a 9B base. 51 vs 49, 21 optimizer steps, LoRA on only 8 of 32 decoder layers and again zero vision tensors. Its declared base `togethercomputer/Qwen3.5-9B` also 404s.
|
| 351 |
- [`manifesta/brandvoice-marketing-model`](https://huggingface.co/manifesta/brandvoice-marketing-model), the Part 1 marketing adapter that did work, 56% win rate on Llama 3.3 70B. Style transfer on a base with a weak default style, which is the case where a small adapter has room to move.
|
| 352 |
|
| 353 |
+
## Everything behind these weights is public
|
| 354 |
+
|
| 355 |
+
The dataset this adapter was trained on, the scripts that built it, and a verifier that rechecks
|
| 356 |
+
every number claimed here against the live artifacts:
|
| 357 |
+
|
| 358 |
+
**<https://github.com/A1VARA5/verified-math-code-17k>**
|
| 359 |
+
|
| 360 |
+
```bash
|
| 361 |
+
git clone https://github.com/A1VARA5/verified-math-code-17k
|
| 362 |
+
cd verified-math-code-17k
|
| 363 |
+
python verify.py
|
| 364 |
+
```
|
| 365 |
+
|
| 366 |
+
Standard library only, no install step and no account. 23 checks, and the same 23 run on a daily
|
| 367 |
+
schedule in GitHub Actions, so the badge above goes red if any claim on this card stops being true.
|
| 368 |
+
The tensor facts in this card are among the checks: layer coverage is re-derived from the published
|
| 369 |
+
`adapter_model.safetensors` by name, not copied from the config.
|
| 370 |
+
|
| 371 |
+
Dataset: <https://huggingface.co/datasets/manifesta/verified-math-code-17k>
|
| 372 |
+
|
| 373 |
+
|
| 374 |
## License
|
| 375 |
|
| 376 |
- **Adapter weights:** Apache-2.0.
|