31 lines
753 B
Plaintext
31 lines
753 B
Plaintext
# ollama-utils-metadata
|
|
# hf_upstream: https://huggingface.co/Qwen/Qwen3-14B-GGUF
|
|
# capabilities: tools,thinking
|
|
FROM ./qwen3-14b-q4_k_m.gguf
|
|
|
|
# Context and output parameters
|
|
PARAMETER num_ctx 40960
|
|
PARAMETER num_predict 32768
|
|
|
|
# Sampling parameters (non-thinking mode)
|
|
PARAMETER temperature 0.7
|
|
PARAMETER top_p 0.8
|
|
PARAMETER top_k 20
|
|
PARAMETER min_p 0.00
|
|
PARAMETER repeat_penalty 1.0
|
|
PARAMETER presence_penalty 1.5
|
|
|
|
# Stop tokens
|
|
PARAMETER stop "<|im_start|>"
|
|
PARAMETER stop "<|im_end|>"
|
|
|
|
# Template for Qwen3 chat format
|
|
TEMPLATE """{{ if .System }}<|im_start|>system
|
|
{{ .System }}<|im_end|>
|
|
{{ end }}<|im_start|>user
|
|
{{ .Prompt }}<|im_end|>
|
|
<|im_start|>assistant
|
|
"""
|
|
|
|
SYSTEM """You are Qwen, created by Alibaba Cloud. You are a helpful assistant."""
|