sampling : add XTC sampler (#9742)
* Initial XTC commit Adds XTC sampler, not activated by default, but recommended settings by default. * Cleanup * Simplified chances calculation To be more inline with the original implementation, chance is calculated once at the beginning. * First fixes by comments Still need to look into sorting * Fixed trailing backspaces * Fixed RNG to be reproduceable Thanks to @slaren for directions * Fixed forgotten header * Moved `min_keep` Moved from conditions to a simple check at the end. * Fixed broken randomization Thanks to @slaren for explanation * Swapped sorting for a custom algorithm Shifts tokens to remove the penalized ones, then puts the penalized at the back. Should make `min_keep` still viable. * Algorithm rework 1. Scan token from top till the first non-penalizable 2. Remove the last captured token (the least probable above threshold) 3. Shift all tokens to override the remaining penalizable 4. Penalize and put them at the the bottom. * Added XTC to `test-sampling` * Simplified algorithm and more tests * Updated info in common and args * Merged back lost commits in common and arg * Update dump info in common * Fixed incorrect min_keep check * Added XTC to README * Renamed parameters, fixed info and defaults * probability is at 0 by default, but XTC is included in sampling queue * threshold higher than 0.5 switches XTC off * Initial server support * Added XTC to server UIs * Fixed labels in old server UI * Made algorithm safer and more readable * Removed xtc_threshold_max * Fixed arg after update * Quick fixes by comments * Simplified algorithm since threshold_max is removed * Renamed random distribution * Fixed tests and outdated README * Small fixes
This commit is contained in:
MaggotHATE
GitHub
parent
dcdd535302
commit
fbc98b748e
11 changed files with 195 additions and 10 deletions
|
|
@ -241,6 +241,19 @@ The `--mirostat-ent` option sets the Mirostat target entropy (tau), which repres
|
|||
|
||||
Example usage: `--mirostat 2 --mirostat-lr 0.05 --mirostat-ent 3.0`
|
||||
|
||||
### XTC Sampling
|
||||
|
||||
- `--xtc-probability N`: Sets the chance for token removal (checked once on sampler start) (default: 0.0).
|
||||
- `--xtc-threshold N`: Sets a minimum probability threshold for tokens to be removed (default: 0.1).
|
||||
|
||||
Exclude Top Choices (XTC) is a unique sampler that is designed to remove top tokens from consideration and avoid more obvious and repetitive outputs. With a chance of `xtc-probability` it searches for tokens with probabilities of `xtc-threshold` and above, then removes all such tokens except the least probable one.
|
||||
|
||||
By removing top tokens XTC can improve the variety of answers, break writing clichés and inhibit repition, since clichés and repeated phrases are usually more likely to appear. By keeping the last token above the threshold, XTC ensures that the answer is still coherent. XTC is meant to be used for creative tasks, but feel free to experiment with different settings for different models.
|
||||
|
||||
Being experimental and unique, XTC is disabled by default. The recommended combination of samplers is Min-P followed by XTC on its default settings: `--sampling-seq mx --min-p 0.02 --xtc-probability 0.5`.
|
||||
|
||||
Example usage: `--xtc-probability 0.5 --xtc-threshold 0.1`
|
||||
|
||||
### Logit Bias
|
||||
|
||||
- `-l TOKEN_ID(+/-)BIAS, --logit-bias TOKEN_ID(+/-)BIAS`: Modify the likelihood of a token appearing in the generated text completion.
|
||||
|
|
|
|||
|
|
@ -43,6 +43,8 @@
|
|||
top_k: 0, // <= 0 to use vocab size
|
||||
top_p: 1.0, // 1.0 = disabled
|
||||
min_p: 0.05, // 0 = disabled; recommended for non-english: ~ 0.4
|
||||
xtc_probability: 0.0, // 0 = disabled;
|
||||
xtc_threshold: 0.1, // > 0.5 disables XTC;
|
||||
tfs_z: 1.0, // 1.0 = disabled
|
||||
typical_p: 1.0, // 1.0 = disabled
|
||||
presence_penalty: 0.0, // 0.0 = disabled
|
||||
|
|
@ -836,6 +838,8 @@ return html`
|
|||
${FloatField({ label: "TFS-Z", title: "Activates tail-free sampling, a method used to limit the prediction of tokens that are too frequent. The parameter z controls the strength of this limitation. A value of 1.0 means that this function is deactivated.", max: 1.0, min: 0.0, name: "tfs_z", step: 0.01, value: params.value.tfs_z })}
|
||||
${FloatField({ label: "Frequency Penalty", title: "A penalty that is applied based on the frequency with which certain tokens occur in the training data set. A higher value results in rare tokens being favoured.", max: 1.0, min: 0.0, name: "frequency_penalty", step: 0.01, value: params.value.frequency_penalty })}
|
||||
${FloatField({ label: "Typical-P", title: "Activates local typical sampling, a method used to limit the prediction of tokens that are atypical in the current context. The parameter p controls the strength of this limitation. A value of 1.0 means that this function is deactivated.", max: 1.0, min: 0.0, name: "typical_p", step: 0.01, value: params.value.typical_p })}
|
||||
${FloatField({ label: "XTC probability", title: "Sets the chance for token removal (checked once on sampler start)", max: 1.0, min: 0.0, name: "xtc_probability", step: 0.01, value: params.value.xtc_probability })}
|
||||
${FloatField({ label: "XTC threshold", title: "Sets a minimum probability threshold for tokens to be removed", max: 0.5, min: 0.0, name: "xtc_threshold", step: 0.01, value: params.value.xtc_threshold })}
|
||||
${IntField({ label: "Min Keep", title: "If greater than 0, samplers are forced to return N possible tokens at minimum. Default is 0", max: 10, min: 0, name: "min_keep", value: params.value.min_keep })}
|
||||
</fieldset>
|
||||
|
||||
|
|
@ -1132,6 +1136,8 @@ document.addEventListener('DOMContentLoaded', (event) => {
|
|||
const snapSettings = {
|
||||
temperature: { snapValue: 1.0, snapRangeMultiplier: 6 },
|
||||
min_p: { snapValue: 0.05, snapRangeMultiplier: 2 },
|
||||
xtc_probability: { snapValue: 0.0, snapRangeMultiplier: 4 },
|
||||
xtc_threshold: { snapValue: 0.5, snapRangeMultiplier: 4 },
|
||||
top_p: { snapValue: 1.0, snapRangeMultiplier: 4 },
|
||||
tfs_z: { snapValue: 1.0, snapRangeMultiplier: 4 },
|
||||
typical_p: { snapValue: 1.0, snapRangeMultiplier: 4 },
|
||||
|
|
|
|||
|
|
@ -307,6 +307,8 @@
|
|||
top_k: 40, // <= 0 to use vocab size
|
||||
top_p: 0.95, // 1.0 = disabled
|
||||
min_p: 0.05, // 0 = disabled
|
||||
xtc_probability: 0.0, // 0 = disabled;
|
||||
xtc_threshold: 0.1, // > 0.5 disables XTC;
|
||||
tfs_z: 1.0, // 1.0 = disabled
|
||||
typical_p: 1.0, // 1.0 = disabled
|
||||
presence_penalty: 0.0, // 0.0 = disabled
|
||||
|
|
@ -1013,6 +1015,8 @@
|
|||
${FloatField({ label: "Typical P", max: 1.0, min: 0.0, name: "typical_p", step: 0.01, value: params.value.typical_p })}
|
||||
${FloatField({ label: "Presence penalty", max: 1.0, min: 0.0, name: "presence_penalty", step: 0.01, value: params.value.presence_penalty })}
|
||||
${FloatField({ label: "Frequency penalty", max: 1.0, min: 0.0, name: "frequency_penalty", step: 0.01, value: params.value.frequency_penalty })}
|
||||
${FloatField({ label: "XTC probability", max: 1.0, min: 0.0, name: "xtc_probability", step: 0.01, value: params.value.xtc_probability })}
|
||||
${FloatField({ label: "XTC threshold", max: 0.5, min: 0.0, name: "xtc_threshold", step: 0.01, value: params.value.xtc_threshold })}
|
||||
</fieldset>
|
||||
<hr />
|
||||
<fieldset class="three">
|
||||
|
|
|
|||
|
|
@ -863,6 +863,8 @@ struct server_context {
|
|||
slot.sparams.top_k = json_value(data, "top_k", default_sparams.top_k);
|
||||
slot.sparams.top_p = json_value(data, "top_p", default_sparams.top_p);
|
||||
slot.sparams.min_p = json_value(data, "min_p", default_sparams.min_p);
|
||||
slot.sparams.xtc_probability = json_value(data, "xtc_probability", default_sparams.xtc_probability);
|
||||
slot.sparams.xtc_threshold = json_value(data, "xtc_threshold", default_sparams.xtc_threshold);
|
||||
slot.sparams.tfs_z = json_value(data, "tfs_z", default_sparams.tfs_z);
|
||||
slot.sparams.typ_p = json_value(data, "typical_p", default_sparams.typ_p);
|
||||
slot.sparams.temp = json_value(data, "temperature", default_sparams.temp);
|
||||
|
|
@ -1196,6 +1198,8 @@ struct server_context {
|
|||
{"top_k", slot.sparams.top_k},
|
||||
{"top_p", slot.sparams.top_p},
|
||||
{"min_p", slot.sparams.min_p},
|
||||
{"xtc_probability", slot.sparams.xtc_probability},
|
||||
{"xtc_threshold", slot.sparams.xtc_threshold},
|
||||
{"tfs_z", slot.sparams.tfs_z},
|
||||
{"typical_p", slot.sparams.typ_p},
|
||||
{"repeat_last_n", slot.sparams.penalty_last_n},
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue