156 Commits

Author SHA1 Message Date
Robert
d6636897ad Update diarize.py 2024-05-04 13:42:47 -07:00
Robert
4e761eb620 Gpt 2024-05-04 11:56:16 -07:00
WalkThroughTheDoorAndDoTheDinosaur
e13d1c02b4 Cleanup 2024-04-30 21:13:59 -07:00
WalkThroughTheDoorAndDoTheDinosaur
9ae51b022e cleanup main folder 2024-04-30 21:08:56 -07:00
WalkThroughTheDoorAndDoTheDinosaur
7249b9a7b3 Updated README instructions and clarified how things work. 2024-04-30 21:07:26 -07:00
WalkThroughTheDoorAndDoTheDinosaur
048b346370 I love python packages 2024-04-30 20:50:07 -07:00
Robert
aa9a2ff806 Create pylint.yml 2024-04-30 09:51:06 -07:00
Mike
764e03bde0 roller with chatgpt json 2023-12-28 15:32:24 -05:00
the-crypt-keeper
bd63a139e1 Merge pull request #1 from the-crypt-keeper/json
v2 wip
2023-12-10 00:35:55 -05:00
Mike
ffe8bac9a3 shake off the dust 2023-12-10 00:35:26 -05:00
Mike
380372cdf5 a few more summaries 2023-08-15 02:16:54 +00:00
Mike
973d7b9802 use assertive tone in speaker extraction prompt 2023-08-15 02:07:43 +00:00
Mike
2ee56b3810 robust speaker detection, two more videos now work 2023-08-15 02:04:22 +00:00
Mike
94ca3d135f working summary from 13b model 2023-08-15 01:38:01 +00:00
Mike
5d43febdce update wip 2023-08-14 20:47:55 -04:00
Mike
52a5db52c0 new form data 2023-08-14 20:47:00 -04:00
Mike
4914cdd89e better debug outputs, track speakers 2023-08-06 12:20:17 -04:00
Mike
be6ed72ca6 working prompt for bhenrym14/airoboros-33b-gpt4-1.4.1-PI-8192-GPTQ at 4k 2023-08-05 19:42:05 -04:00
Mike
10f539c2ba updates 2023-08-05 19:29:33 -04:00
Mike
a466607d09 improve summary quality, but still crashing at 30/36 on lex 2023-08-05 15:17:08 -04:00
Mike
898f937064 tweak summary prompt to maintain third person, got further then ever 2023-08-05 14:26:04 -04:00
Mike
39d65052c3 looser json parsing (allow newline in strings) 2023-08-05 13:32:03 -04:00
Mike
2e9444b426 update summary generation prompt and remove common answer prefixes 2023-08-05 12:32:47 -04:00
Mike
d198eda582 improved prompt, generate more tokens 2023-08-05 12:00:34 -04:00
Mike
0df0c15ca4 seems to work! 2023-08-05 11:46:09 -04:00
Mike
8c36cbd916 airoboros 33b 2.0 roller 2023-08-05 10:20:45 -04:00
Mike
976348774e split roller from chunker 2023-08-05 09:26:23 -04:00
Mike
3a80c005ad working v2 chunker 2023-08-05 09:15:03 -04:00
Mike
3e812863ff dialogue offset feature 2023-08-04 12:38:25 -04:00
Mike
39bcf38b53 handle wack characters in video filenames 2023-08-04 12:24:33 -04:00
Mike
89ba7d8590 works! 2023-08-04 12:09:53 -04:00
Mike
b00e6779ae updates 2023-08-04 11:41:07 -04:00
Mike
df95e86878 diarize cli wip 2023-08-04 11:26:55 -04:00
Mike
2a31092ca8 wip 2023-08-04 10:46:46 -04:00
Mike
8a2ae4b90c sorta works 2023-08-04 01:07:22 -04:00
Mike
cae837d18d kinda works 2023-08-04 00:42:18 -04:00
Mike
e800b65b71 run the 4 testcases 2023-08-02 19:07:53 -04:00
Mike
4d90ac164e switch to json instead of text so we can chunk by time 2023-08-02 18:45:21 -04:00
Mike
632992e32f improve readme 2023-07-31 11:41:15 -04:00
Mike
7e5ac58d3a remove chunker outputs, you can rebuild them with the repo itself 2023-07-30 15:44:16 -04:00
Mike
ef9ece02d3 more ufo compares 2023-07-30 15:43:54 -04:00
Mike
7b473b550a fixes 2023-07-30 14:29:33 -04:00
Mike
c4f008519b updates 2023-07-30 14:28:53 -04:00
Mike
70b3bafd82 aoe results 2023-07-30 14:28:47 -04:00
Mike
ca89524af5 readme wip 2023-07-30 13:11:35 -04:00
Mike
3841d5e8ed generic script 2023-07-30 13:05:40 -04:00
Mike
b57ef286ce consitent naming 2023-07-30 13:05:26 -04:00
Mike
2d2b973e3d reorg 2023-07-30 12:47:54 -04:00
Mike
89910e2b92 add youtube link 2023-07-30 11:50:58 -04:00
Mike
963ed20ae4 wut 2023-07-30 11:49:57 -04:00