tigerllm-1b-gguf

🐯 TigerLLM-1B GGUF (F16 + Q4_K_M)

---

🧠 Model Overview

TigerLLM-1B is a 1B parameter language model converted into GGUF format for efficient local inference and Bangla text generation.

• Base Model: md-nishat-008/TigerLLM-1B-it
• Format: GGUF (llama.cpp compatible)
• Precision Options:
  • F16 → High quality, high memory usage
  • Q4_K_M → Fast, low memory, recommended
• Use Case: Offline CPU inference and strong output in bangla compare to other models.
• Key Feature: Lightweight multilingual inference (including Bangla testing)

---

⚙️ Why llama.cpp Instead of Ollama

• quantized(Q4_K_M) format will run on ollama no problem, but we prefer to run in llama.cpp cause,
• llama.cpp supports latest GGUF formats and experimental models
• Ollama can reject newer GGUF metadata or architectures
• llama.cpp is more flexible and up-to-date for custom models