---
license: odc-by
language: [en]
library_name: transformers
pipeline_tag: text-generation
tags: [legal, finance, llama, sft, instruction-tuned]
---

# slm125MLIVE-sft

Instruction fine-tuned (SFT) version of `thesreedath/slm-125m-base` (125M, LLaMA).
Trained on ~8,000 grounded Q&A pairs (RAFT-style: answer from the provided
context) synthesized with Gemini 2.5 Flash and grounding-judged.

- SFT val perplexity: ~2.71
- Format: chat with `<|system|> <|user|> <|assistant|>` special tokens.

## Usage
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("sumitguha13/slm125MLIVE-sft")
model = AutoModelForCausalLM.from_pretrained("sumitguha13/slm125MLIVE-sft")
prompt = ("<|bos|><|system|>\nYou are a helpful assistant. Answer using only the "
          "provided context.\n<|user|>\nContext:\n<PASSAGE>\n\nQuestion: <Q>\n<|assistant|>\n")
ids = tok(prompt, return_tensors="pt", add_special_tokens=False).input_ids
print(tok.decode(model.generate(ids, max_new_tokens=120, repetition_penalty=1.3)[0], skip_special_tokens=True))
```

Small base model: use a context for grounded answers; generations may be imperfect.