RIFA-PRO
RIFA Intelligence For All — Pro (~3B)
Created & maintained by SM Shahbaj
About
| Property | Value |
|---|---|
| Name | RIFA-PRO |
| Parameters | ~3B |
| Creator | SM Shahbaj |
| Languages | English, Bangla (Bengali) |
| License | Apache 2.0 |
RIFA-PRO is the Pro-tier model in the RIFA family — bilingual chat (English + Bangla), fine-tuned and released by SM Shahbaj.
Downloads
Safetensors
| Format | Link |
|---|---|
| Merged weights | Browse files |
GGUF (LM Studio / Ollama / llama.cpp)
| Format | Link |
|---|---|
| F16 | Download |
| Q8_0 | Download |
| Q6_K | Download |
| Q5_K_M ⭐ | Download |
| Q4_K_M | Download |
| Q3_K_M | Download |
⭐ Recommended: Q5_K_M · Low VRAM: Q4_K_M / Q3_K_M
Quick start
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
repo = "smshahbaj/RIFA-Pro-3B"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, torch_dtype=torch.bfloat16, device_map="auto")
messages = [
{"role": "system", "content": "You are RIFA-PRO, created by SM Shahbaj."},
{"role": "user", "content": "Who are you?"},
]
text = tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tok(text, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=128)
print(tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
LM Studio system prompt
You are RIFA-PRO, an AI assistant created by SM Shahbaj. Answer helpfully in English or Bangla as appropriate.
Ollama
FROM ./RIFA-PRO-3B.Q5_K_M.gguf
SYSTEM "You are RIFA-PRO, created by SM Shahbaj."
Limitations
- ~3B class model — strong for size, not frontier-scale
- May hallucinate; verify important facts
- For consistent RIFA branding, use the system prompt above
SM Shahbaj · huggingface.co/smshahbaj · Apache 2.0
Documentation
- Series hub (GitHub): https://github.com/smshahbaj/RIFA-Text-Series
- Model page: RIFA-PRO docs
- HF collection: https://huggingface.co/collections/smshahbaj/rifa-text-series