PEFT
Safetensors
tiny-aya-global
json-extraction
nested-json
structured-output
aligntune
curatorkit
auditkit
Instructions to use sujalgawas/aya-json-sftV1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use sujalgawas/aya-json-sftV1 with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("CohereLabs/tiny-aya-global") model = PeftModel.from_pretrained(base_model, "sujalgawas/aya-json-sftV1") - Notebooks
- Google Colab
- Kaggle
sujalgawas/aya-json-sftV1 β Stage-2 Hierarchical & Nested JSON Extraction SFT
Second-stage QLoRA adapter fine-tuned on top of sujalgawas/aya-json-sft (base model CohereLabs/tiny-aya-global) using a 50% Stage-1 Flat + 50% Stage-2 Hierarchical Nested dataset (sujalgawas/aya-json-sftV1-dataset).
Stage-1 vs. Stage-2 Nested Benchmark Results
| Model | Split | Schema Valid | Doc Exact | Nested Object Acc | Array Completeness | Null Acc | Hallucination Rate |
|---|---|---|---|---|---|---|---|
Stage-1 (sujalgawas/aya-json-sft) |
test_nested |
0.0 | 0.0 | 0.0626 | 0.0456 | 0.0578 | 0.9091 |
Stage-2 (sujalgawas/aya-json-sftV1) |
test_nested |
0.7143 | 0.7143 | 0.8288 | 0.7868 | 0.8776 | 0.1122 |
Stage-2 (sujalgawas/aya-json-sftV1) |
test_flat |
1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 0.0 |
- Downloads last month
- 25
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support