Commit Graph

518 Commits

Author SHA1 Message Date
Webifi 40e38d8b54 change default delimiter 2023-09-01 20:49:20 -05:00
Webifi ddd17798e7 Merge delimiter with last known buffer 2023-09-01 19:24:15 -05:00
Webifi 4d27497bc7 Simplify Llama-chat format for now 2023-09-01 19:22:52 -05:00
Webifi 01f7a657db Change default request config for Llama instruct 2023-09-01 19:14:27 -05:00
Webifi 4f18d40000 Add base llama-65b model 2023-09-01 18:06:04 -05:00
Webifi 69b107e8f7 remove Llama-2-70b-hf model 2023-09-01 18:04:23 -05:00
Webifi 872e6c8566 open new websocket if model changes 2023-09-01 17:42:35 -05:00
Webifi 47f96c0bb9 update chat settings update on each request 2023-09-01 17:42:16 -05:00
Webifi 6fc236d7f0 profile tweak 2023-08-31 20:25:37 -05:00
Webifi b4780dc036 attempt to fix websocket timeouts 2023-08-31 20:25:19 -05:00
Webifi 57f6423f13 missed some reins 2023-08-31 18:44:33 -05:00
Webifi a41a3af243 Simplify jenLama. Remove all reigns. Faster. 2023-08-31 18:13:05 -05:00
Webifi 2e4181bf7e Add holdWebsocket option for faster petals chat 2023-08-31 17:25:16 -05:00
Webifi e991dfd9b7 filter settings on URL import/export 2023-08-31 04:47:51 -05:00
Webifi cd0cd9ebef add hideSystemPrompt 2023-08-31 04:28:59 -05:00
Webifi 0d4f12d900 Apply profile changes from URI before start 2023-08-29 13:55:22 -05:00
Webifi 189e0acfb1 Add option to copy profile as URL 2023-08-29 13:50:50 -05:00
Webifi 5c3fac4cbe fix #273 2023-08-28 15:34:26 -05:00
Webifi 784d60c3d6 Fix missing tweaks from rollback 2023-08-26 18:22:34 -05:00
Webifi db0beb8c51 fix profile issue 2023-08-26 18:12:41 -05:00
Webifi 462cfbd2fd enable Llama-2 non-chat model 2023-08-25 16:54:17 -05:00
Webifi 334d90ffc0 continue to tweak profile dialog 2023-08-25 15:56:56 -05:00
Webifi 57c5a5574a Update save/send shortcut to ctrl-enter 2023-08-25 11:27:53 -05:00
Webifi a8e7ad0ff6 Keep loading prompt after summary 2023-08-25 00:24:26 -05:00
Webifi d4e12b49fb Keep loading prompt after summary 2023-08-25 00:13:43 -05:00
Webifi 57bb9d0922 Profile tweak 2023-08-24 21:58:50 -05:00
Webifi 58508efe57 Profile tweak 2023-08-24 21:54:43 -05:00
Webifi 7e44a63fec alt-s to save and send message 2023-08-24 19:48:03 -05:00
Webifi a278d54a15 Take top_p down a touch 2023-08-24 18:13:12 -05:00
Webifi b3a71dd3b0 Simplify prompts now that we have repetitionPenalty 2023-08-22 16:21:09 -05:00
Webifi 85c5587a9f Set new chat settings via URI 2023-08-20 15:44:05 -05:00
Webifi 86211b0910 Reduce repetitionPenalty from recommended 1.2 to 1.1 2023-08-20 14:08:44 -05:00
Webifi 3593744035 typo 2023-08-20 14:01:01 -05:00
Webifi 754f8b5d69 Change default repetitionPenalty to 1.2 2023-08-20 13:57:07 -05:00
Webifi fc46a21ab7 fix typo 2023-08-20 13:49:33 -05:00
Webifi cd443a6e06 Use repititionPenalty 2023-08-20 13:19:38 -05:00
Webifi 70c2f5928a Enable repitition_penalty setting for petals 2023-08-20 13:11:20 -05:00
Webifi cf0fa4d5fc Add summary prompt 2023-08-19 21:15:28 -05:00
Webifi 0e1bdfe9a4 prompt tweak 2023-08-19 20:36:02 -05:00
Webifi 3b72f73080 fix prompt continuation for petals 2023-08-19 19:16:18 -05:00
Webifi 54aac40788 Allow resume of truncated petals completions 2023-08-19 18:52:53 -05:00
Webifi 8ce425722b tweak profile 2023-08-19 17:56:59 -05:00
Webifi f805f08f98 Add ability for non-submitted notes in system prompt 2023-08-19 16:26:24 -05:00
Webifi 96696347a2 tweak profile 2023-08-19 15:57:26 -05:00
Webifi deb75832e7 specify profile for new chat via uri 2023-08-19 13:09:43 -05:00
Webifi ed0dbe1188 Enable petals via uri 2023-08-19 13:09:13 -05:00
Webifi 98dbaec35a Fix missing prompt prefix 2023-08-19 12:44:37 -05:00
Webifi 43792b1d69 Update CheapGPT 2023-08-18 13:25:33 -05:00
Webifi 5eea1c8ddd fix typo 2023-08-18 13:14:20 -05:00
Webifi f9183c0662 Shrink text-areas on profile change 2023-08-18 13:06:51 -05:00
Webifi c70b2c3928 auto apply system prompt changes to current chat 2023-08-18 12:47:47 -05:00
Webifi 538e9d749a Enable system prompt for CheapGPT 2023-08-18 12:35:53 -05:00
Webifi ff97a30e78 tweak profile 2023-08-17 21:34:05 -05:00
Webifi a260d49d5a Add more variation in response 2023-08-17 21:32:29 -05:00
Webifi 726842179c typo 2023-08-17 18:46:12 -05:00
Webifi e5f6be8f81 typos 2023-08-17 18:27:10 -05:00
Webifi bc7ba1da74 Try to make jenLlama profile more creative in responses 2023-08-17 18:20:28 -05:00
Webifi 8edccda1eb Allow more leeway in name suggestion token count 2023-08-17 12:24:39 -05:00
Webifi 950a27d8e6 Add more debug, fix non-streaming response 2023-08-17 12:20:09 -05:00
Webifi e00aad20f5 Force model type for jenLama profile 2023-08-17 08:22:41 -05:00
Webifi 083e31d4a7 Add uninhibited LLaMA profile 2023-08-17 07:58:43 -05:00
Webifi 5f919098f5 Add CheapGPT profile 2023-08-17 07:58:04 -05:00
Webifi f523f8d4bc Fix typo 2023-08-16 15:20:57 -05:00
Webifi cb2b9e07f4 Add distinction between chat and instruct models 2023-08-16 15:20:07 -05:00
Webifi f4d9774423 update token count 2023-08-16 01:51:49 -05:00
Webifi 86f427f62f Fix potential infinite loop 2023-08-15 23:48:49 -05:00
Webifi a08d8bcd54 Move token counting to model detail. 2023-08-15 21:46:33 -05:00
Webifi 91885384a1 disable non-chat Llama-2 2023-08-15 20:39:19 -05:00
Webifi fb2290308f Begin refactoring model providers to be less anti-pattern 2023-08-15 20:32:30 -05:00
Webifi af568efd3a Add lead prompts 2023-08-07 12:44:16 -05:00
Webifi bcb2b93e84 Default aggressive stop on 2023-08-07 12:37:57 -05:00
Webifi 7c588ce212 Add StableBeluga2. Update prompt structures 2023-08-07 12:33:42 -05:00
Webifi 8b2f2515f9 Send summary request as system prompt 2023-08-07 12:32:46 -05:00
Webifi c473d731ce Properly close websocket connections. 2023-07-28 18:10:45 -05:00
Webifi bb2a177f22 remove debug 2023-07-28 16:48:36 -05:00
Webifi 4c37969635 remove debug 2023-07-28 16:48:07 -05:00
Webifi 0f12fcdb95 Format LLama-2-chat prompts to spec. 2023-07-28 16:46:50 -05:00
Webifi f223f4e510 Fix disabled OpenAI models #243 2023-07-26 17:49:26 -05:00
Webifi 58afe8f375 typo 2023-07-25 14:55:07 -05:00
Webifi 1ef08110c3 Add aggressive stop setting 2023-07-25 14:53:29 -05:00
Webifi b0812796a1 Add link to swarm health 2023-07-25 13:28:14 -05:00
Webifi ff3799637b Allow scrolling while streaming re: #241 2023-07-25 00:21:04 -05:00
Webifi 0ffdd78863 Another prompt prefix issue 2023-07-24 23:14:28 -05:00
Webifi 833633991a Fix user prompt prefix injection 2023-07-24 22:20:52 -05:00
Webifi af08f5c99e Update text 2023-07-24 20:00:20 -05:00
Webifi 190bf16ce6 Drop llama 65b and guanaco 65b - unstable in swarm 2023-07-24 19:56:05 -05:00
Webifi 38d38bf948 Fix some issues with stop sequences and role sequences 2023-07-24 19:48:28 -05:00
Webifi f56e29b829 Show shorter model name 2023-07-24 15:53:04 -05:00
Webifi f6380e1cc2 Allow Petals and/or OpenAI 2023-07-24 15:26:17 -05:00
Webifi ca19bab19d Don't allow too low of temp or top_p 2023-07-22 17:21:01 -05:00
Webifi 7aadca3c5c Better error handling for Petals 2023-07-22 17:08:40 -05:00
Webifi 15dcd27e8f Get temp and top_p working for Petals 2023-07-22 16:48:26 -05:00
Webifi 6d35a46d50 Allow user to adjust message chaining strings 2023-07-22 14:40:12 -05:00
Webifi 9a6004c55d More changes for Petals integration 2023-07-22 13:24:18 -05:00
Webifi df222e7028 Try to import chat name suggestion 2023-07-22 13:23:24 -05:00
Webifi cdbd17fa13 Don't show generation setting that don't apply 2023-07-22 08:56:12 -05:00
Webifi 914055f1f9 Initial test of Petals as alternative to OpenAI 2023-07-22 01:42:21 -05:00
Webifi 8e35b198da Add [[LAST-PROMPT]] system prompt expansion 2023-07-20 20:32:36 -05:00
Webifi b12ec15d80 Use "the user" in system prompts, not "I" 2023-07-17 21:11:11 -05:00
Webifi 0b45a1833b Update profile example 2023-07-17 20:03:33 -05:00