Instructions to use Puujeeeeeeeeeeee/sft-task-patterns-v2-adapter with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Puujeeeeeeeeeeee/sft-task-patterns-v2-adapter with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Puujeeeeeeeeeeee/sft-task-patterns-v2-adapter", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Unsloth Desktop
Upload TRAINING_METADATA.md with huggingface_hub
Browse files- TRAINING_METADATA.md +10 -0
TRAINING_METADATA.md
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Puujeeeeeeeeeeee/sft-task-patterns-v2-adapter
|
| 2 |
+
|
| 3 |
+
- run_id: sft_task_patterns_wholerow_20260803_074411
|
| 4 |
+
- instruction_version: wholerow-v2-no-decomposition
|
| 5 |
+
- design: whole-row, no decomposition (7-10x past cpt-round4's real CPT ceiling -- accepted trade-off)
|
| 6 |
+
- training_data_path: /workspace/mn_synthesis_data_patterns_wholerow_v1/synthesis_training_data.jsonl
|
| 7 |
+
- training_data_sha256_12: 84253cb93972
|
| 8 |
+
- train_examples: 144
|
| 9 |
+
- base_model: Puujeeeeeeeeeeee/cpt-round5
|
| 10 |
+
- trained_at: 2026-08-03T08:04:14.825182
|