Update default textgen presets
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.68, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.17, |
| 4 | 5 | "top_k": 77, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0.42, |
| 7 | 7 | "tfs": 0.97, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.02, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.7, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.92, |
| 4 | 5 | "top_k": 150, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | 10 | "rep_pentypical_p": 4.51, |
| 11 | + "min_p": 0, | |
| 12 | + "rep_pen": 3, | |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 2, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 10, |
| 15 | 19 | "length_penalty": 1.4, |
| 16 | 20 | "min_length": 200, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": true, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.87, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.99, |
| 4 | 5 | "top_k": 85, |
| 5 | - "typical_p": 0.68, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 0.68, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 0.68, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.01, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 4, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0.6, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": false, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.7, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.5, |
| 4 | 5 | "top_k": 40, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.2, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0, |
| 4 | 5 | "top_k": 1, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.18, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": false, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.31, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.14, |
| 4 | 5 | "top_k": 49, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0.52, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 1.49, |
| 9 | 9 | "eta_cutoff": 10.42, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.17, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.7, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.5, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 0.19, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 0.19, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.66, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 0.6, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 0.6, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.7, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.1, |
| 4 | 5 | "top_k": 40, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.18, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 200, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.98, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.37, |
| 4 | 5 | "top_k": 100, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.18, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,24 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.06, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 2, |
| 21 | 42 | "mirostat_tau": 9.61, |
| 22 | 43 | "mirostat_eta": 1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 23 | 84 | "rep_pen_size": 0 |
| 24 | 85 | } |
| @@ -1,24 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.17, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 2, |
| 21 | 42 | "mirostat_tau": 9.91, |
| 22 | 43 | "mirostat_eta": 1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 23 | 84 | "rep_pen_size": 0 |
| 24 | 85 | } |
| @@ -1,24 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.17, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 2, |
| 21 | 42 | "mirostat_tau": 9.62, |
| 22 | 43 | "mirostat_eta": 1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 23 | 84 | "rep_pen_size": 0 |
| 24 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 2, |
| 21 | 42 | "mirostat_tau": 8, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.7, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.85, |
| 4 | 5 | "top_k": 50, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.8, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.9, |
| 4 | 5 | "top_k": 100, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.15, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.99, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 100, |
| 5 | - "typical_p": 0.97, | |
| 6 | - "rep_pen": 1, | |
| 7 | 6 | "top_a": 0, |
| 8 | 7 | "tfs": 1, |
| 9 | 8 | "epsilon_cutoff": 0, |
| 10 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 0.97, | |
| 11 | + "min_p": 0, | |
| 12 | + "rep_pen": 1, | |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.63, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.98, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.05, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.99, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.85, |
| 4 | 5 | "top_k": 12, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.15, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.07, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 100, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.05, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.44, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.15, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.99, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.18, |
| 4 | 5 | "top_k": 30, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.15, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.72, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.73, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.53, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.64, |
| 4 | 5 | "top_k": 33, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0.04, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.07, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.7, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.9, |
| 4 | 5 | "top_k": 20, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.15, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.31, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.29, |
| 4 | 5 | "top_k": 72, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.09, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.02, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.95, |
| 4 | 5 | "top_k": 50, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.7, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 1, |
| 4 | 5 | "top_k": 0, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0.2, |
| 7 | 7 | "tfs": 0.95, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.15, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 1.01, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.21, |
| 4 | 5 | "top_k": 91, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0.75, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 10.78, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.21, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1.07, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -11,6 +11,8 @@ | ||
| 11 | 11 | "min_p": 0.1, |
| 12 | 12 | "rep_pen": 1, |
| 13 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 14 | 16 | "no_repeat_ngram_size": 0, |
| 15 | 17 | "penalty_alpha": 0, |
| 16 | 18 | "num_beams": 1, |
| @@ -19,24 +21,56 @@ | ||
| 19 | 21 | "encoder_rep_pen": 1, |
| 20 | 22 | "freq_pen": 0, |
| 21 | 23 | "presence_pen": 0, |
| 24 | + "skew": 0, | |
| 22 | 25 | "do_sample": true, |
| 23 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 24 | 38 | "add_bos_token": true, |
| 25 | - "truncation_length": 2048, | |
| 26 | 39 | "ban_eos_token": false, |
| 27 | 40 | "skip_special_tokens": true, |
| 28 | - "streaming": true, | |
| 29 | 41 | "mirostat_mode": 0, |
| 30 | 42 | "mirostat_tau": 5, |
| 31 | 43 | "mirostat_eta": 0.1, |
| 32 | 44 | "guidance_scale": 1, |
| 33 | 45 | "negative_prompt": "", |
| 34 | 46 | "grammar_string": "", |
| 47 | + "json_schema": {}, | |
| 35 | 48 | "banned_tokens": "", |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 36 | 71 | "ignore_eos_token": false, |
| 37 | 72 | "spaces_between_special_tokens": true, |
| 38 | 73 | "typespeculative_ngram": "ooba"false, |
| 39 | - "legacy_api": false, | |
| 40 | 74 | "sampler_order": [ |
| 41 | 75 | 5, |
| 42 | 76 | 6, |
| @@ -46,5 +80,6 @@ | ||
| 46 | 80 | 3, |
| 47 | 81 | 4 |
| 48 | 82 | ], |
| 83 | + "logit_bias": [], | |
| 49 | 84 | "rep_pen_size": 0 |
| 50 | 85 | } |
| @@ -11,6 +11,8 @@ | ||
| 11 | 11 | "min_p": 0.1, |
| 12 | 12 | "rep_pen": 1, |
| 13 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 14 | 16 | "no_repeat_ngram_size": 0, |
| 15 | 17 | "penalty_alpha": 0, |
| 16 | 18 | "num_beams": 1, |
| @@ -19,24 +21,56 @@ | ||
| 19 | 21 | "encoder_rep_pen": 1, |
| 20 | 22 | "freq_pen": 0, |
| 21 | 23 | "presence_pen": 0, |
| 24 | + "skew": 0, | |
| 22 | 25 | "do_sample": true, |
| 23 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 24 | 38 | "add_bos_token": true, |
| 25 | - "truncation_length": 2048, | |
| 26 | 39 | "ban_eos_token": false, |
| 27 | 40 | "skip_special_tokens": true, |
| 28 | - "streaming": true, | |
| 29 | 41 | "mirostat_mode": 0, |
| 30 | 42 | "mirostat_tau": 5, |
| 31 | 43 | "mirostat_eta": 0.1, |
| 32 | 44 | "guidance_scale": 1, |
| 33 | 45 | "negative_prompt": "", |
| 34 | 46 | "grammar_string": "", |
| 47 | + "json_schema": {}, | |
| 35 | 48 | "banned_tokens": "", |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 36 | 71 | "ignore_eos_token": false, |
| 37 | 72 | "spaces_between_special_tokens": true, |
| 38 | 73 | "typespeculative_ngram": "ooba"false, |
| 39 | - "legacy_api": false, | |
| 40 | 74 | "sampler_order": [ |
| 41 | 75 | 5, |
| 42 | 76 | 6, |
| @@ -46,5 +80,6 @@ | ||
| 46 | 80 | 3, |
| 47 | 81 | 4 |
| 48 | 82 | ], |
| 83 | + "logit_bias": [], | |
| 49 | 84 | "rep_pen_size": 0 |
| 50 | 85 | } |
| @@ -11,6 +11,8 @@ | ||
| 11 | 11 | "min_p": 0.1, |
| 12 | 12 | "rep_pen": 1, |
| 13 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 14 | 16 | "no_repeat_ngram_size": 0, |
| 15 | 17 | "penalty_alpha": 0, |
| 16 | 18 | "num_beams": 1, |
| @@ -19,24 +21,56 @@ | ||
| 19 | 21 | "encoder_rep_pen": 1, |
| 20 | 22 | "freq_pen": 0, |
| 21 | 23 | "presence_pen": 0, |
| 24 | + "skew": 0, | |
| 22 | 25 | "do_sample": true, |
| 23 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 24 | 38 | "add_bos_token": true, |
| 25 | - "truncation_length": 2048, | |
| 26 | 39 | "ban_eos_token": false, |
| 27 | 40 | "skip_special_tokens": true, |
| 28 | - "streaming": true, | |
| 29 | 41 | "mirostat_mode": 0, |
| 30 | 42 | "mirostat_tau": 5, |
| 31 | 43 | "mirostat_eta": 0.1, |
| 32 | 44 | "guidance_scale": 1, |
| 33 | 45 | "negative_prompt": "", |
| 34 | 46 | "grammar_string": "", |
| 47 | + "json_schema": {}, | |
| 35 | 48 | "banned_tokens": "", |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 36 | 71 | "ignore_eos_token": false, |
| 37 | 72 | "spaces_between_special_tokens": true, |
| 38 | 73 | "typespeculative_ngram": "ooba"false, |
| 39 | - "legacy_api": false, | |
| 40 | 74 | "sampler_order": [ |
| 41 | 75 | 5, |
| 42 | 76 | 6, |
| @@ -46,5 +80,6 @@ | ||
| 46 | 80 | 3, |
| 47 | 81 | 4 |
| 48 | 82 | ], |
| 83 | + "logit_bias": [], | |
| 49 | 84 | "rep_pen_size": 0 |
| 50 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.82, |
| 3 | + "temperature_last": true, | |
| 3 | 4 | "top_p": 0.21, |
| 4 | 5 | "top_k": 72, |
| 5 | - "typical_p": 1, | |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | + "typical_p": 1, | |
| 11 | + "min_p": 0, | |
| 10 | 12 | "rep_pen": 1.19, |
| 11 | 13 | "rep_pen_range": 0, |
| 14 | + "rep_pen_decay": 0, | |
| 15 | + "rep_pen_slope": 1, | |
| 12 | 16 | "no_repeat_ngram_size": 0, |
| 13 | 17 | "penalty_alpha": 0, |
| 14 | 18 | "num_beams": 1, |
| 15 | 19 | "length_penalty": 1, |
| 16 | 20 | "min_length": 0, |
| 17 | 21 | "encoder_rep_pen": 1, |
| 22 | + "freq_pen": 0, | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 18 | 25 | "do_sample": true, |
| 19 | 26 | "early_stopping": false, |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 20 | 41 | "mirostat_mode": 0, |
| 21 | 42 | "mirostat_tau": 5, |
| 22 | 43 | "mirostat_eta": 0.1, |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 5, | |
| 76 | + 6, | |
| 77 | + 0, | |
| 78 | + 1, | |
| 79 | + 2, | |
| 80 | + 3, | |
| 81 | + 4 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 23 | 85 | } |
| @@ -1,23 +1,85 @@ | ||
| 1 | 1 | { |
| 2 | 2 | "temp": 0.65, |
| 3 | 3 | "top_ptemperature_last": 0.47true, |
| 4 | 4 | "top_ktop_p": 420.47, |
| 5 | 5 | "typical_ptop_k": 142, |
| 6 | 6 | "top_a": 0, |
| 7 | 7 | "tfs": 1, |
| 8 | 8 | "epsilon_cutoff": 0, |
| 9 | 9 | "eta_cutoff": 0, |
| 10 | 10 | "rep_pentypical_p": 1.18, |
| 11 | 11 | "rep_pen_rangemin_p": 0, |
| 12 | 12 | "no_repeat_ngram_sizerep_pen": 01.18, |
| 13 | 13 | "penalty_alpharep_pen_range": 0, |
| 14 | 14 | "num_beamsrep_pen_decay": 10, |
| 15 | 15 | "length_penaltyrep_pen_slope": 1, |
| 16 | 16 | "min_lengthno_repeat_ngram_size": 0, |
| 17 | 17 | "encoder_rep_penpenalty_alpha": 10, |
| 18 | 18 | "do_samplenum_beams": true1, |
| 19 | 19 | "early_stoppinglength_penalty": false1, |
| 20 | 20 | "mirostat_modemin_length": 0, |
| 21 | 21 | "mirostat_tauencoder_rep_pen": 51, |
| 22 | 22 | "mirostat_etafreq_pen": 0.1 , |
| 23 | -} | |
| 23 | + "presence_pen": 0, | |
| 24 | + "skew": 0, | |
| 25 | + "do_sample": true, | |
| 26 | + "early_stopping": false, | |
| 27 | + "dynatemp": false, | |
| 28 | + "min_temp": 0, | |
| 29 | + "max_temp": 2, | |
| 30 | + "dynatemp_exponent": 1, | |
| 31 | + "smoothing_factor": 0, | |
| 32 | + "smoothing_curve": 1, | |
| 33 | + "dry_allowed_length": 2, | |
| 34 | + "dry_multiplier": 0, | |
| 35 | + "dry_base": 1.75, | |
| 36 | + "dry_sequence_breakers": "[\"\\n\", \":\", \"\\\"\", \"*\"]", | |
| 37 | + "dry_penalty_last_n": 0, | |
| 38 | + "add_bos_token": true, | |
| 39 | + "ban_eos_token": false, | |
| 40 | + "skip_special_tokens": true, | |
| 41 | + "mirostat_mode": 0, | |
| 42 | + "mirostat_tau": 5, | |
| 43 | + "mirostat_eta": 0.1, | |
| 44 | + "guidance_scale": 1, | |
| 45 | + "negative_prompt": "", | |
| 46 | + "grammar_string": "", | |
| 47 | + "json_schema": {}, | |
| 48 | + "banned_tokens": "", | |
| 49 | + "sampler_priority": [ | |
| 50 | + "temperature", | |
| 51 | + "dynamic_temperature", | |
| 52 | + "quadratic_sampling", | |
| 53 | + "top_k", | |
| 54 | + "top_p", | |
| 55 | + "typical_p", | |
| 56 | + "epsilon_cutoff", | |
| 57 | + "eta_cutoff", | |
| 58 | + "tfs", | |
| 59 | + "top_a", | |
| 60 | + "min_p", | |
| 61 | + "mirostat" | |
| 62 | + ], | |
| 63 | + "samplers": [ | |
| 64 | + "top_k", | |
| 65 | + "tfs_z", | |
| 66 | + "typical_p", | |
| 67 | + "top_p", | |
| 68 | + "min_p", | |
| 69 | + "temperature" | |
| 70 | + ], | |
| 71 | + "ignore_eos_token": false, | |
| 72 | + "spaces_between_special_tokens": true, | |
| 73 | + "speculative_ngram": false, | |
| 74 | + "sampler_order": [ | |
| 75 | + 6, | |
| 76 | + 0, | |
| 77 | + 1, | |
| 78 | + 3, | |
| 79 | + 4, | |
| 80 | + 2, | |
| 81 | + 5 | |
| 82 | + ], | |
| 83 | + "logit_bias": [], | |
| 84 | + "rep_pen_size": 0 | |
| 85 | +} | |