Sample Output

#2
by Dunjeon - opened

I'm currently testing some samplers for better adherence, but this is strong.

Adherence

Snip

{"temperature", 0.7},
{"top_p", 0.75},
{"top_k", 30},
{"top_a", 0},
{"tfs", 1},
{"typical_p", 1},
{"min_p", 0.02},
{"rep_pen", 1.1},
{"rep_pen_range", 2048},
{"rep_pen_decay", 0},
{"rep_pen_slope", 1},
{"presence_pen", 0.03},
{"dynatemp", false}, //found it better without, but you can flip it on and see how it goes
{"min_temp", 0.7},
{"max_temp", 0.9},
{"dynatemp_exponent", 0.85},
{"smoothing_factor", 0.25},
{"smoothing_curve", 1},
{"dry_allowed_length", 2},
{"dry_multiplier", 0.8},
{"dry_base", 1.75},
{"dry_penalty_last_n", 4096},
{"sampler_order", new List<int> { 6, 0, 1, 3, 4, 2, 5 }},
...

Sign up or log in to comment