Woof 2B v1.1 GGUF

Woof 2B v1.1 (Terrier) in GGUF, for llama.cpp. It is the chat model in Underdog for Android, the same model the iPhone app runs with MLX.

Model details

• Source: ConwayResearch/woof-2B-mlx-bf16-v1.1, converted with llama.cpp b11485's convert_hf_to_gguf.py, then quantized with llama-quantize.
• Parameters: 2,516,756,480 (about 2.52 billion). Llama architecture.
• File: woof-2B-1.1-Q4_K_M.gguf, 1,561,318,432 bytes, SHA-256 e173464b652feab53855ab0038d5b8cf825429385274cae9a8ab060c634ba65f.
• The chat template is embedded in the GGUF.

Usage

llama-cli -m woof-2B-1.1-Q4_K_M.gguf -cnv

Other formats

MLX 4-bit and MLX BF16.

License

CC BY-NC 4.0 (see LICENSE). You can use, share and adapt it for non-commercial purposes with credit to Underdog. Commercial use needs a separate license: write to license@conway.tech.

Built from woof-2B-mlx-bf16-v1.1. Upstream components keep their original licenses (see LICENSE-APACHE).