More tweaks

Author: mattjcly

en/config.json

@@ -99,14 +99,14 @@
   "llm.prediction.tools/serverPageDescriptionAddon": "Pass this through the request body as `tools` when using the server API",
   "llm.prediction.promptTemplate/title": "Prompt Template",
   "llm.prediction.promptTemplate/subTitle": "The format in which messages in chat are sent to the model. Changing this may introduce unexpected behavior - make sure you know what you're doing!",
-  "llm.prediction.mlx.speculativeDecoding.numDraftTokensExact/title": "Draft Tokens to Generate",
-  "llm.prediction.mlx.speculativeDecoding.numDraftTokensExact/subTitle": "The number of tokens to generate with the draft model per main model token. Find the sweet spot of compute vs. reward",
-  "llm.prediction.llama.speculativeDecoding.minContinueDraftingProbability/title": "Drafting Probability Cutoff",
-  "llm.prediction.llama.speculativeDecoding.minContinueDraftingProbability/subTitle": "Continue drafting until a token's probability falls below this threshold. Higher values generally mean lower risk, lower reward",
-  "llm.prediction.llama.speculativeDecoding.minDraftLengthToConsider/title": "Min Draft Size",
-  "llm.prediction.llama.speculativeDecoding.minDraftLengthToConsider/subTitle": "Drafts smaller than this will be ignored by the main model. Higher values generally mean lower risk, lower reward",
-  "llm.prediction.llama.speculativeDecoding.maxTokensToDraft/title": "Max Draft Size",
-  "llm.prediction.llama.speculativeDecoding.maxTokensToDraft/subTitle": "Max number of tokens allowed in a draft. Ceiling if all token probs are > the cutoff. Lower values generally mean lower risk, lower reward",
+  "llm.prediction.speculativeDecoding.numDraftTokensExact/title": "Draft Tokens to Generate",
+  "llm.prediction.speculativeDecoding.numDraftTokensExact/subTitle": "The number of tokens to generate with the draft model per main model token. Find the sweet spot of compute vs. reward",
+  "llm.prediction.speculativeDecoding.minContinueDraftingProbability/title": "Drafting Probability Cutoff",
+  "llm.prediction.speculativeDecoding.minContinueDraftingProbability/subTitle": "Continue drafting until a token's probability falls below this threshold. Higher values generally mean lower risk, lower reward",
+  "llm.prediction.speculativeDecoding.minDraftLengthToConsider/title": "Min Draft Size",
+  "llm.prediction.speculativeDecoding.minDraftLengthToConsider/subTitle": "Drafts smaller than this will be ignored by the main model. Higher values generally mean lower risk, lower reward",
+  "llm.prediction.speculativeDecoding.maxTokensToDraft/title": "Max Draft Size",
+  "llm.prediction.speculativeDecoding.maxTokensToDraft/subTitle": "Max number of tokens allowed in a draft. Ceiling if all token probs are > the cutoff. Lower values generally mean lower risk, lower reward",
   "llm.prediction.speculativeDecoding.draftModel/title": "Draft Model",
   "llm.prediction.reasoning.parsing/title": "Reasoning Section Parsing",
   "llm.prediction.reasoning.parsing/subTitle": "How to parse reasoning sections in the model's output",