115 Commits

Author	SHA1	Message	Date
somebody	bce1a907e5	Update aux device to depend on primary device	2023-07-03 19:36:31 -05:00
somebody	d4b923a054	Remove debug	2023-06-21 17:41:15 -05:00
somebody	947bcc58e4	Experiments	2023-06-21 17:33:14 -05:00
somebody	1546b9efaa	Hello its breaking breakmodel time	2023-05-27 16:31:53 -05:00
Llama	3768848548	Fix tokenization and whitespace issues with llama-derived models ... Work around the 'soft' prefix space behavior of sentencepiece. Override encode to restore the deleted HF support for decode_with_prefix_space. Override decode to skip the soft space and return true decoded tokens. Allow submitting chat messages with embedded newlines. Split sentences between punctuation and whitespace, rather than after whitespace. Also include trailing quotes and brackets after sentence stoppers. This avoids splitting ." and .) into two tokens, for instance. Insert whitespace at the beginning of the author's note, since sentences are split with leading whitespace. Remove spurious newlines at the end of chat responses.	2023-05-03 01:27:11 -07:00
somebody	f6b5548131	Support safetensors in get_sharded_checkpoint_num_tensors	2023-05-01 19:15:27 -05:00
somebody	97e84928ba	Download all shards correctly on aria2 and raise on bad load key	2023-05-01 18:53:36 -05:00
somebody	f9fb5eba89	Remove debug	2023-04-15 18:56:49 -05:00
somebody	5dd67d027a	Workaround for socketio context errors for loading	2023-04-15 18:54:21 -05:00
somebody	a2ae87d1b7	Utils: Support safetensors aria2 download	2023-04-15 11:51:16 -05:00
somebody	65b60085e3	Undo debug	2023-03-13 20:30:46 -05:00
somebody	adc11fdbc9	TPUMTJ: Fix loading bar ... I don't know why it works but I know it works	2023-03-13 20:13:05 -05:00
somebody	ef1155291f	Model: TPU Fixes	2023-02-28 18:05:34 -06:00
somebody	a73804ca1e	Accelerate: Remove HAS_ACCELERATE ... Accelerate has been a dependency for a while, and as such we probably shouldn't be lugging around code that assumes it isn't present.	2023-02-26 12:18:06 -06:00
somebody	ffe4f25349	Model: Work on stoppers and stuff	2023-02-25 17:12:16 -06:00
somebody	6b4905de30	Model: Port rest of models over ... Generation's still broke but it's a start	2023-02-25 16:05:56 -06:00
somebody	f8c4158ebc	Model: Successful load implementation ... The goal of this series of commits is to have an implementation-agnostic interface for models, thus being less reliant on HF Transformers for model support. A model object will have a method for generation, a list of callbacks to be run on every token generation, a list of samplers that will modify probabilities, etc. Basically anything HF can do should be easily implementable with the new interface :^) Currently I've tested the loading of pre-downloaded models with breakmodel between GPUs and that works, though essentially no testing has been done in the larger scheme of things. Currently this is about the only supported configuration, and generation isn't very functional.	2023-02-24 21:41:44 -06:00
Henk	cd566caf20	Revision Fixes (Removes the workaround)	2023-02-19 00:51:50 +01:00
Henk	a9a724e38c	Merge branch 'main' into united	2023-02-18 18:14:03 +01:00
Henk	f57489f73c	Revision Cleanup	2023-01-31 18:46:59 +01:00
Henk	257a535be5	Revision Fixes Fixes	2023-01-31 05:17:34 +01:00
Henk	739cccd8ed	Revision Fixes	2023-01-31 04:48:46 +01:00
Henk	640bd64037	Revision Fixes (And Var Workaround)	2023-01-31 04:00:35 +01:00
Henk	f1739dd184	Chatmode Regex	2023-01-13 13:04:54 +01:00
ebolam	db2fbbb17a	Colab Debug	2022-10-24 19:16:03 -04:00
ebolam	b01713875c	Add status bar message and status bar for TPU loading	2022-10-24 18:34:45 -04:00
ebolam	e101e43132	Merge commit 'refs/pull/243/head' of https://github.com/ebolam/KoboldAI into UI2	2022-10-23 12:03:40 -04:00
ebolam	06ee1426a1	Progress Bar for UI1 fixed ... Console print for model loading fixed	2022-10-22 21:05:07 -04:00
ebolam	fbc7a3ff3d	Chat mode cleanup	2022-10-13 10:46:42 -04:00
ebolam	ea738180de	Un-pegging Transformer version	2022-10-07 15:18:40 -04:00
vfbd	6758d5b538	Merge branch 'united' into mkultra	2022-09-28 14:30:34 -04:00
ebolam	f442c1c9b9	Merge pull request #150 from henk717/united ... Update to united	2022-09-21 13:50:29 -04:00
Henk	fef946c173	Possible colab aria2 status fix	2022-09-21 19:48:17 +02:00
Henk	06f4d9addf	No Aria2 spam	2022-09-21 19:36:57 +02:00
ebolam	09ae3d7c84	Merge commit 'refs/pull/149/head' of https://github.com/ebolam/KoboldAI into UI2	2022-09-21 13:35:32 -04:00
ebolam	8915ee7eb3	Fix for aria2 download status to UI	2022-09-21 13:09:27 -04:00
Henk	cca3ce3493	Aria2 Fixes	2022-09-21 18:57:09 +02:00
Henk	f62c740f7e	Revert "Aria2 Fixes" ... This reverts commit 8d1c734df8.	2022-09-21 18:47:13 +02:00
Henk	8d1c734df8	Aria2 Fixes	2022-09-21 18:21:48 +02:00
ebolam	4d8df761ba	Colab Fix	2022-09-19 09:45:38 -04:00
ebolam	1a31e0994c	Colab Fix	2022-09-19 09:40:49 -04:00
ebolam	cb5204731c	Colab Fix	2022-09-19 09:38:03 -04:00
ebolam	2d058afe3b	Colab Fix	2022-09-19 09:37:31 -04:00
ebolam	cd9e4175a9	Colab Fix	2022-09-19 09:31:37 -04:00
ebolam	48101f492d	Colab Fix	2022-09-19 09:30:30 -04:00
ebolam	d1c56bc6b8	Colab Fix	2022-09-19 09:28:33 -04:00
ebolam	f5ce00622b	Fix for UI2 download status bar	2022-09-19 09:21:08 -04:00
ebolam	90b5dab721	Fix	2022-09-19 08:05:08 -04:00
ebolam	8e26918752	merge fix	2022-09-18 20:34:21 -04:00
ebolam	fc5a5952f0	Fix	2022-09-18 20:31:21 -04:00