Datasets:
Tasks:
Text Generation
Languages:
English
Size:
n<1K
Tags:
code
notebooks
training-scripts
dataset:Nanthasit/sakthai-kaggle-notebooks
license-mit
dataset-card
License:
List new 0.5B train + BFCL eval scripts in Contents
Browse files
README.md
CHANGED
|
@@ -28,6 +28,8 @@ tags:
|
|
| 28 |
|---|---|
|
| 29 |
| `sakthai-engine.ipynb` | Training notebook (1.5B) |
|
| 30 |
| `sakthai-7b-engine.ipynb` | Training notebook (7B) |
|
|
|
|
|
|
|
| 31 |
| `scripts/sakthai-7b-post-train.py` | LoRA merge → GGUF convert → upload |
|
| 32 |
| `deploy-endpoint.py` | Deploy to HF Inference Endpoints |
|
| 33 |
|
|
|
|
| 28 |
|---|---|
|
| 29 |
| `sakthai-engine.ipynb` | Training notebook (1.5B) |
|
| 30 |
| `sakthai-7b-engine.ipynb` | Training notebook (7B) |
|
| 31 |
+
| `train-sakthai-0.5b-v2.py` | Improved 0.5B fine-tune (MLP LoRA targets, r16/rsLoRA, completion-only loss, v6 + irrelevance) |
|
| 32 |
+
| `eval-bfcl-0.5b.py` | BFCL-style eval on the held-out `test` split (simple / parallel / irrelevance) |
|
| 33 |
| `scripts/sakthai-7b-post-train.py` | LoRA merge → GGUF convert → upload |
|
| 34 |
| `deploy-endpoint.py` | Deploy to HF Inference Endpoints |
|
| 35 |
|