49 lines
1.2 KiB
Markdown
49 lines
1.2 KiB
Markdown
---
|
|
license: apache-2.0
|
|
language:
|
|
- en
|
|
pipeline_tag: text-generation
|
|
tags:
|
|
- unsloth
|
|
- lora
|
|
- sft
|
|
- dpo
|
|
- extraction
|
|
- wingit
|
|
---
|
|
|
|
# WingIt Extractor v1
|
|
|
|
## Model Description
|
|
|
|
This is a specialized model for **extracting high-value dialogue, concepts, and principles from dating/pickup transcripts**.
|
|
|
|
It was trained using:
|
|
- **SFT**: 1,372 instruction examples (Prompt 3)
|
|
- **DPO**: Preference optimization for correct classification
|
|
|
|
## Use Cases
|
|
|
|
- Extracting dialogues from raw transcripts
|
|
- Identifying behavioral concepts (frame control, abundance, push-pull, etc.)
|
|
- Classifying content as dialogue/concept/principle
|
|
- Generating structured JSONL training data
|
|
|
|
## Training Data
|
|
|
|
- `Prompt 3_cleaned.jsonl`: 1,372 instruction examples
|
|
- Base model: `unsloth/llama-3.1-8b-instruct-bnb-4bit`
|
|
|
|
## Usage
|
|
|
|
```python
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
model = AutoModelForCausalLM.from_pretrained("richierich007/wingit-extractor-v1")
|
|
tokenizer = AutoTokenizer.from_pretrained("richierich007/wingit-extractor-v1")
|
|
|
|
prompt = "Extract dialogue and concepts from this transcript..."
|
|
inputs = tokenizer(prompt, return_tensors="pt")
|
|
outputs = model.generate(**inputs)
|
|
print(tokenizer.decode(outputs[0]))
|