22 10 hours ago

Reasoning model focused on shorter, token-efficient thinking + DFlash2 + map-k4v

f94d19d8c43b · 347B
{
"draft_ngram_map_k4v_min_hits": 1,
"draft_ngram_map_k4v_size_m": 48,
"draft_ngram_map_k4v_size_n": 24,
"draft_num_predict": 7,
"draft_spec_type": "draft-dflash,ngram-map-k4v",
"min_p": 0,
"num_batch": 512,
"num_ctx": 131072,
"num_predict": -1,
"presence_penalty": 0,
"repeat_penalty": 1,
"stop": [
"<|im_start|>",
"<|im_end|>"
],
"temperature": 0.6,
"top_k": 20,
"top_p": 0.95
}