From 0f077968c0b175d4ceeaf9c82e8239f55530ff4c Mon Sep 17 00:00:00 2001 From: Yazan Agha-Schrader Date: Mon, 27 May 2024 23:46:18 +0200 Subject: [PATCH] add tooltips to the parameters with comprehensible explanations --- examples/server/themes/mount-ai/index.html | 38 +++++++++++----------- 1 file changed, 19 insertions(+), 19 deletions(-) diff --git a/examples/server/themes/mount-ai/index.html b/examples/server/themes/mount-ai/index.html index 03f9c2639..a4a33aeb1 100644 --- a/examples/server/themes/mount-ai/index.html +++ b/examples/server/themes/mount-ai/index.html @@ -559,7 +559,7 @@ function updateSystemPrompt(e) { } } - const FloatField = ({ label, max, min, name, step, value }) => { + const FloatField = ({ label, title, max, min, name, step, value }) => { return html`
@@ -579,7 +579,7 @@ function updateSystemPrompt(e) { ` }; - const BoolField = ({ label, name, value }) => { + const BoolField = ({ label, title, name, value }) => { return html`
@@ -793,41 +793,41 @@ function updateSystemPrompt(e) { ${session.value.type === 'chat' ? ChatConfigForm() : CompletionConfigForm()}
- ${IntField({ label: "Prediction", max: 2048, min: -1, step: 16, name: "n_predict", value: params.value.n_predict, })} - ${FloatField({ label: "Min-P sampling", max: 1.0, min: 0.0, name: "min_p", step: 0.01, value: params.value.min_p })} - ${FloatField({ label: "Repetition Penalty", max: 2.0, min: 0.0, name: "repeat_penalty", step: 0.01, value: params.value.repeat_penalty })} - ${FloatField({ label: "Temperature", max: 2.0, min: 0.0, name: "temperature", step: 0.01, value: params.value.temperature })} + ${IntField({ label: "Prediction", title: "Set the maximum number of tokens to predict when generating text. Note: May exceed the set limit slightly if the last token is a partial multibyte character. When 0, no tokens will be generated but the prompt is evaluated into the cache. The value -1 means infinity. Default is 0.8", max: 2048, min: -1, step: 16, name: "n_predict", value: params.value.n_predict, })} + ${FloatField({ label: "Min-P sampling", title: "The minimum probability for a token to be considered, relative to the probability of the most likely token. Note that it's good practice to disable all other samplers aside from temperature when using min-p. It is also recommenend to go this approach. Default is 0.05 – But consider higher values like ~ 0.4 for non-English text generation. The value 1.0 means disabled", max: 1.0, min: 0.0, name: "min_p", step: 0.01, value: params.value.min_p })} + ${FloatField({ label: "Repetition Penalty", title: "Control the repetition of token sequences in the generated text. Default is 1.1", max: 2.0, min: 0.0, name: "repeat_penalty", step: 0.01, value: params.value.repeat_penalty })} + ${FloatField({ label: "Temperature", title: "This will adjust the overall randomness of the generated text. It is the most common sampler. Default is 0.8 but consider using lower values for more factual texts or for non-English text generation", max: 2.0, min: 0.0, name: "temperature", step: 0.01, value: params.value.temperature })}
Further Options
- ${IntField({ label: "Top-K", max: 100, min: 0, step: 1, name: "top_k", value: params.value.top_k })} - ${IntField({ label: "Penalize Last N", max: 2048, min: 0, step: 16, name: "repeat_last_n", value: params.value.repeat_last_n })} - ${FloatField({ label: "Top-P", max: 1.0, min: 0.0, name: "top_p", step: 0.01, value: params.value.top_p })} - ${FloatField({ label: "Presence Penalty", max: 1.0, min: 0.0, name: "presence_penalty", step: 0.01, value: params.value.presence_penalty })} - ${FloatField({ label: "TFS-Z", max: 1.0, min: 0.0, name: "tfs_z", step: 0.01, value: params.value.tfs_z })} - ${FloatField({ label: "Frequency Penalty", max: 1.0, min: 0.0, name: "frequency_penalty", step: 0.01, value: params.value.frequency_penalty })} - ${FloatField({ label: "Typical-P", max: 1.0, min: 0.0, name: "typical_p", step: 0.01, value: params.value.typical_p })} + ${IntField({ label: "Top-K", title: "Limits the selection of the next token to the K most probable tokens. 1 means no randomness = greedy sampling. If set to 0, it means the entire vocabulary size is considered.", max: 100, min: 0, step: 1, name: "top_k", value: params.value.top_k })} + ${IntField({ label: "Penalize Last N", title: "The last n tokens that are taken into account to penalise repetitions. A value of 0 means that this function is deactivated and -1 means that the entire size of the context is taken into account.", max: 2048, min: 0, step: 16, name: "repeat_last_n", value: params.value.repeat_last_n })} + ${FloatField({ label: "Top-P", title: "Limits the selection of the next token to a subset of tokens whose combined probability reaches a threshold value P = top-P. If set to 1, it means the entire vocabulary size is considered.", max: 1.0, min: 0.0, name: "top_p", step: 0.01, value: params.value.top_p })} + ${FloatField({ label: "Presence Penalty", title: "A penalty that is applied if certain tokens appear repeatedly in the generated text. A higher value leads to fewer repetitions.", max: 1.0, min: 0.0, name: "presence_penalty", step: 0.01, value: params.value.presence_penalty })} + ${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 })} ${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 })}
- - - + + +
- ${FloatField({ label: "Entropy tau", max: 10.0, min: 0.0, name: "mirostat_tau", step: 0.01, value: params.value.mirostat_tau })} - ${FloatField({ label: "Learning-rate eta", max: 1.0, min: 0.0, name: "mirostat_eta", step: 0.01, value: params.value.mirostat_eta })} + ${FloatField({ label: "Entropy tau", title: "Tau controls the desired level of entropy (or 'surprise') in the text. A low tau (e.g. 0.5) would mean that a text is very predictable, but will also be very coherent. A high tau (e.g. 8.0) would mean that the text is very creative and surprising, but may also be difficult to follow because unlikely words will occur frequently.", max: 10.0, min: 0.0, name: "mirostat_tau", step: 0.01, value: params.value.mirostat_tau })} + ${FloatField({ label: "Learning-rate eta", title: "Eta determines how quickly the Mirostat algorithm adjusts its predictions to achieve the desired entropy. A learning rate that is too high can cause the algorithm to react too quickly and possibly become unstable, because the algorithm will try to maintain a balance between surprises and precision in the context of only a few words. In this way, 'the common thread' could be lost. Whereas a learning rate that is too low means that the algorithm reacts too slowly and a red thread becomes a heavy goods train that takes a long time to come to a halt and change a 'topic station'.", max: 1.0, min: 0.0, name: "mirostat_eta", step: 0.01, value: params.value.mirostat_eta })}
- ${IntField({ label: "Show Probabilities", title: "Tooltip-Text hier", max: 10, min: 0, step: 1, name: "n_probs", value: params.value.n_probs })} + ${IntField({ label: "Show Probabilities", title: "If greater than 0, the response also contains the probabilities of top N tokens for each generated token given the sampling settings. The tokens will be colored in gradient from green to red depending on their probabilities. Note that for temperature < 0 the tokens are sampled greedily but token probabilities are still being calculated via a simple softmax of the logits without considering any other sampler settings. Defaults to 0", max: 10, min: 0, step: 1, name: "n_probs", title: "", value: params.value.n_probs })}