29 Commits

Author SHA1 Message Date
Sotirios Anagnostidis 6d569a50dd choice for empty eval 2023-02-11 13:21:02 +01:00
Sotirios Anagnostidis 714677b5d8 add possible kwargs to datasets 2023-02-11 10:44:03 +01:00
Sotirios Anagnostidis 44ed44e05d deactivtae samples mixing by default 2023-02-11 10:33:25 +01:00
theblackcat102 9e69117ead [fix] Fix other PR merge bug 2023-02-11 01:53:37 +00:00
theblackcat102 bcebbbc49c [merge] Fix conflict 2023-02-11 00:23:25 +00:00
sanagnos 4dd0d67e9c Merge pull request #1398 from jackapbutler/fix-tokenizer-match
Add tests and update docstring to tokenizer matching
2023-02-10 11:37:00 +01:00
jack.butler 2fbf2fa457 add docstring info about tokenizer matching 2023-02-10 09:46:58 +00:00
jack.butler 090c5cbcc2 fix tokenizer matching and add tests 2023-02-09 18:47:38 +00:00
sanagnos 4ba622de8e Merge branch 'main' into sft-data-sampling 2023-02-09 09:19:17 +01:00
theblackcat102 a1b90bf981 Merge branch 'main' into add-dataset 2023-02-09 01:28:42 +00:00
Mark Worrall 9faae250ce minor tidy-up 2023-02-08 20:57:13 +00:00
Mark Worrall 283df8ec84 Get working on multi-gpu 2023-02-08 20:49:25 +00:00
Mark Worrall e2caf53654 First version of single GPU sampling working 2023-02-08 08:01:04 +00:00
jack.butler eb1c4ada2a rename arg name to model_name 2023-02-07 09:19:24 +00:00
jack.butler dc7f255f01 add tokenizer config classes 2023-02-06 17:58:55 +00:00
theblackcat102 8b2080559c [fix] Custom collate_fn for training 2023-02-03 06:08:01 +00:00
theblackcat102 f5b2a34857 [feature] add pythia and limit translation pair 2023-01-22 00:56:17 +00:00
theblackcat102 6f6c590e57 [fix] Disable task specific evaluation 2023-01-14 06:47:21 +00:00
theblackcat102 1546111094 [feature] added GSM8k and code refactoring 2023-01-14 06:24:47 +00:00
Sotirios Anagnostidis c8f47eef9f precommits 2023-01-11 22:58:17 +01:00
Sotirios Anagnostidis 4a3ea0b033 refactoring, now running 2023-01-11 22:42:04 +01:00
ekurtulus 5b77dd2e9f better 2023-01-11 11:37:27 +03:00
Sotirios Anagnostidis d3952354e2 pre commits 2023-01-06 22:09:24 +01:00
Sotirios Anagnostidis 88ee3b3264 merge deepspeed 2023-01-06 21:28:26 +01:00
Sotirios Anagnostidis ef02693ac9 quantization 2023-01-06 18:24:28 +01:00
Sotirios Anagnostidis dfaa00dccc gptj 8bit 2023-01-05 00:33:16 +01:00
Sotirios Anagnostidis c20dfaad5b pre-commits 2023-01-03 22:45:34 +01:00
Sotirios Anagnostidis 3a10e9412d Question-Answer special tokens 2023-01-03 22:02:32 +01:00
Sotirios Anagnostidis 675b0866b0 SFT training 2023-01-03 01:31:56 +01:00