Commit Graph

157 Commits

Author SHA1 Message Date
Zach Nussbaum
b1e361882d fix: multi-turn data breaks 2023-04-12 03:51:29 +00:00
Zach Nussbaum
60155de2a6 chore: peft 2023-04-12 03:50:54 +00:00
Zach Nussbaum
3c11e68ec8 feat: commits for eval + generation 2023-04-11 19:14:29 +00:00
Zach Nussbaum
70d75951d1 chore: remove not needed 2023-04-11 12:39:07 +00:00
Zach Nussbaum
84de913baf chore: submodule ff 2023-04-10 02:16:05 +00:00
Zach Nussbaum
f2cb38e65a Merge branch 'gptj' of github.com:nomic-ai/gpt4all into gptj 2023-04-10 02:15:47 +00:00
Zach Nussbaum
6e84421f3f fix: num training steps for lr decay 2023-04-10 02:15:31 +00:00
Zach
8733576664 feat: evals on new gptj models 2023-04-10 02:14:20 +00:00
Zach
5caed073ee fix: topic model for embeddings 2023-04-09 15:12:49 +00:00
Zach Nussbaum
c11ea7dc31 fix: bs try one more time? 2023-04-08 21:47:07 +00:00
Zach Nussbaum
2827693a57 fix: smaller bs for 40gb 2023-04-08 21:36:20 +00:00
Zach Nussbaum
f5ca8bedcf fix: config for lora gptj 2023-04-08 21:17:12 +00:00
Zach Nussbaum
6de58dd1fc fix: saving name 2023-04-08 20:56:13 +00:00
Zach Nussbaum
305fe3d444 Merge remote-tracking branch 'origin/mosaic' into gptj 2023-04-08 20:47:01 +00:00
Zach Nussbaum
c1ca081cda fix: eos/pad token + wd 2023-04-08 20:38:10 +00:00
Zach Nussbaum
5de765410a fix: add wd + min lr to config 2023-04-08 20:37:51 +00:00
Zach Nussbaum
c0a9065032 fix: tokenization error 2023-04-08 20:33:51 +00:00
Zach Nussbaum
df29c62521 fix: config + ignore pkl 2023-04-08 20:33:02 +00:00
Zach Nussbaum
f479a37818 feat: build map script 2023-04-08 19:30:53 +00:00
Zach
2034f0c479 fix: embeddings instead of logits!!! 2023-04-08 17:05:40 +00:00
zanussbaum
fbb788e451 feat: lora gptj 2023-04-07 17:53:07 -04:00
zanussbaum
dcc5204352 fix: batch size 2023-04-07 17:41:45 -04:00
zanussbaum
aa8dd7a636 feat: update for mosaic 2023-04-07 16:54:29 -04:00
Zach Nussbaum
1b5e660476 fix: pyarrow filter 2023-04-07 19:04:19 +00:00
Zach Nussbaum
f974ca651c fix: inference save shards 2023-04-07 16:23:34 +00:00
Zach
573272ad69 fix: drop uneven batch size 2023-04-07 12:09:31 +00:00
Zach
d93f6c2ced fix: concat 2023-04-07 04:33:34 +00:00
Zach
57eb786756 fix: data for inference 2023-04-07 01:45:07 +00:00
Zach
3f9fea47d7 feat: inference for embedding plots 2023-04-07 01:40:39 +00:00
Zach
8f2eb1a583 fix: grad accum loss calc 2023-04-06 12:11:10 +00:00
Zach
e4e88dff33 fix: data processing 2023-04-06 03:03:34 +00:00
Zach
c2fc164779 fix: configs 2023-04-05 20:42:35 +00:00
Zach
838b19bea5 fix: try except push 2023-04-05 20:42:22 +00:00
Zach Nussbaum
885b7f1a3a feat: multinode setup 2023-04-05 02:53:04 +00:00
Zach Nussbaum
98ceac34e1 fix: gptj multinode 2023-04-05 02:52:44 +00:00
Zach Nussbaum
0c8666f635 fix: ignore env 2023-04-05 02:52:21 +00:00
Zach Nussbaum
9567f445b8 fix: only on first process, not once on every node 2023-04-05 02:36:22 +00:00
Zach Nussbaum
ad33b83a48 fix: eval func 2023-04-04 23:25:37 +00:00
Zach
8dd99cc00a fix: prompt len for larger 2023-04-04 22:01:55 +00:00
Zach Nussbaum
63ff39653d feat: gpt-j config 2023-04-04 20:58:08 +00:00
Zach Nussbaum
32357c920f feat: adamw, fix training, log gradients 2023-04-04 20:57:42 +00:00
Zach Nussbaum
c68311810a fix: clean up data, pad at end 2023-04-04 20:53:23 +00:00
Zach Nussbaum
f45eb001a1 fix: clean where prompt is randomly 1 char 2023-04-04 20:47:21 +00:00
Zach Nussbaum
c7f4ca25e4 chore: gitignore ckpts 2023-04-04 20:46:57 +00:00
Andriy Mulyar
9d30a6063e Merge pull request #3 from nomic-ai/train
log wandb multi-epoch
2023-03-29 13:50:26 -04:00
Andriy Mulyar
0eef27364d Qualified number of epochs for LoRa weights 2023-03-29 12:26:47 -04:00
Andriy Mulyar
d61f0077ee Merge pull request #42 from tiendung/main
Change git clone url to https://github.com/nomic-ai/gpt4all.git to avoid `Permission denied`
2023-03-29 10:50:44 -04:00
Andriy Mulyar
4ebac17592 Added Torrent Magnet Link 2023-03-29 10:47:19 -04:00
Andriy Mulyar
f1af332b22 Merge pull request #36 from Hello1024/patch-1
Update README.md to add torrent link to data
2023-03-29 10:38:26 -04:00
Andriy Mulyar
1e0eb13ce7 Merge branch 'main' into patch-1 2023-03-29 10:38:17 -04:00