h34v7's picture
Update README.md
e40dc78 verified
metadata
license: apache-2.0
language:
  - en
  - ru
base_model:
  - h34v7/DXP-Zero-V1.0-24b-Small-Instruct

DXP-Zero-V1.0-24b-Small-Instruct-i1-GGUF

BF16 available here.

Recommended Settings

"temperature": 0.8,
"top_k": 40,
"top_p": 0.95,
"min_p": 0.05,
"repeat_last_n": 40,
"repeat_penalty": 1.2,

Run on Ollama

These are non-imatrix. I'll release the imatrix version later.

GGUF 3-bit Q3_K_M about 27 GB of vRAM/RAM:

ollama run hf.co/h34v7/DXP-Zero-V1.0-24b-Small-Instruct-i1-GGUF:Q3_K_M

GGUF 4-bit Q4_K_M about 30 GB of vRAM/RAM:

ollama run hf.co/h34v7/DXP-Zero-V1.0-24b-Small-Instruct-i1-GGUF:Q4_K_M

GGUF 5-bit Q5_K_M about 33 GB of vRAM/RAM:

ollama run hf.co/h34v7/DXP-Zero-V1.0-24b-Small-Instruct-i1-GGUF:Q5_K_M