Commit Graph

  • 5daf7d5299 Merge pull request #273 from OpenAccess-AI-Collective/NanoCode012-patch-1 NanoCode012 2023-07-14 21:09:50 +09:00
  • 5491278a79 Feat: Add save_safetensors NanoCode012 2023-07-14 13:21:47 +09:00
  • 1514739f0f Set push to hub as private by default NanoCode012 2023-07-14 13:17:49 +09:00
  • 896c1aebcf Feat(docs): Add model_revision arg NanoCode012 2023-07-14 12:56:07 +09:00
  • ef17e15483 Merge pull request #272 from OpenAccess-AI-Collective/model-revision Wing Lian 2023-07-13 23:12:00 -04:00
  • 69a235061b support for loading a model by git revision Wing Lian 2023-07-13 22:58:25 -04:00
  • f6721baf10 tweak to make it work when we have no explicit test split openorca-v2 openorca Wing Lian 2023-07-11 22:40:21 -04:00
  • 0f2a16aa33 use different perplexity calc compute-perplexity-metrics Wing Lian 2023-07-10 13:43:50 -04:00
  • e7c84254ba fix perplexity calculation and make it configurable Wing Lian 2023-07-08 13:35:24 -04:00
  • 1d02606934 compute perplexity from cross entropy Wing Lian 2023-07-08 12:14:54 -04:00
  • 687d889928 Merge pull request #271 from OpenAccess-AI-Collective/quadratic-warmup Wing Lian 2023-07-10 12:48:02 -04:00
  • c4cf567b55 Merge branch 'main' into quadratic-warmup Wing Lian 2023-07-10 12:42:12 -04:00
  • c49729d2bc better configuration for quadratic warmup Wing Lian 2023-07-10 11:52:59 -04:00
  • 13ac4d8de2 Merge pull request #268 from OpenAccess-AI-Collective/fix-adam-args Wing Lian 2023-07-08 12:33:34 -04:00
  • 19cf0bda99 params are adam_*, not adamw_* Wing Lian 2023-07-08 12:13:39 -04:00
  • f74edd5b56 Merge pull request #266 from OpenAccess-AI-Collective/trust-remote-no-llama Wing Lian 2023-07-07 21:38:11 -04:00
  • d69da99c2c skip explicit model type too if using trust_remote_code Wing Lian 2023-07-07 21:33:11 -04:00
  • 66afb76a15 don't use llama if trust_remote_code is set since that needs to use AutoModel path Wing Lian 2023-07-07 21:31:02 -04:00
  • a692ad3f4c Merge pull request #264 from OpenAccess-AI-Collective/NanoCode012-patch-1 NanoCode012 2023-07-06 23:34:57 +09:00
  • 41da98b982 Fix for linter NanoCode012 2023-07-06 23:20:11 +09:00
  • 9e64f42e0f Fix local path loading and custom strategy type NanoCode012 2023-07-06 23:08:09 +09:00
  • b9b7d4ce92 Merge pull request #221 from utensil/local_dataset Wing Lian 2023-07-03 09:10:13 -04:00
  • 9bed281867 Merge pull request #258 from NanoCode012/fix/deprecate-push Wing Lian 2023-07-03 09:08:26 -04:00
  • e79c8e617e Fix future deprecation push_to_hub_model_id NanoCode012 2023-07-03 12:44:29 +09:00
  • 71456955f5 pin pydantic so deepspeed isn't broken Wing Lian 2023-07-02 22:26:51 -04:00
  • 33814cc94e make sure we eval for openorca Wing Lian 2023-07-02 17:59:10 -04:00
  • 50254a7ccc handle orca splits Wing Lian 2023-07-01 07:20:23 -04:00
  • 3a783c04e4 Merge pull request #247 from OpenAccess-AI-Collective/fix-apex-base Wing Lian 2023-07-01 06:18:25 -04:00
  • 1e5014acec Merge pull request #255 from OpenAccess-AI-Collective/open-orca-prompts Wing Lian 2023-07-01 01:11:23 -04:00
  • a10da1caff 11.7.0 nvidia/cuda docker images are deprecated, move to 11.7.1 dev-base Wing Lian 2023-07-01 00:29:07 -04:00
  • 4066c78631 Merge pull request #246 from OpenAccess-AI-Collective/sys-prompts-instruct Wing Lian 2023-07-01 00:27:29 -04:00
  • 78a1e1fa12 open orca support Wing Lian 2023-07-01 00:19:41 -04:00
  • bc8a2e5547 Merge pull request #249 from OpenAccess-AI-Collective/NanoCode012-patch-1 NanoCode012 2023-06-30 15:01:41 +09:00
  • 910ebe47f5 Merge pull request #252 from OpenAccess-AI-Collective/NanoCode012-readme-fix NanoCode012 2023-06-30 14:56:55 +09:00
  • c146880a75 Update README.md NanoCode012 2023-06-30 11:33:53 +09:00
  • 77bdb7d144 Fix typing list NanoCode012 2023-06-29 14:29:55 +09:00
  • 530809fd74 update pip install command for apex Wing Lian 2023-06-28 22:36:28 -04:00
  • 924bbfddec add option for instruct w sys prompts Wing Lian 2023-06-28 22:27:17 -04:00
  • f150c027e3 Merge pull request #224 from OpenAccess-AI-Collective/system-prompt-data Wing Lian 2023-06-27 17:57:43 -04:00
  • 5c39c006c9 Merge pull request #244 from OpenAccess-AI-Collective/push-to-hub Wing Lian 2023-06-27 17:57:30 -04:00
  • 612aabd8c4 push intermediate model checkpoints to hub Wing Lian 2023-06-27 15:40:25 -04:00
  • af05883f75 Merge pull request #243 from OpenAccess-AI-Collective/unprompted-instruct Wing Lian 2023-06-25 22:50:35 -04:00
  • 05ab9092e3 skip the system prompt Wing Lian 2023-06-25 22:40:50 -04:00
  • 7b57ed7618 pylint for duplicated code for system prompts Wing Lian 2023-06-18 06:40:28 -04:00
  • 3a38271276 add tests and supoort for loader for sys prompt data Wing Lian 2023-06-17 23:52:40 -04:00
  • 8d20e0a3d3 initial wip to get sys prompt from dataset Wing Lian 2023-06-17 19:22:58 -04:00
  • de8ed229c3 Merge pull request #240 from OpenAccess-AI-Collective/tokenizer-fast Wing Lian 2023-06-25 12:47:55 -04:00
  • 478d8c7b8e Merge pull request #241 from OpenAccess-AI-Collective/py3-pre-commit Wing Lian 2023-06-25 12:47:02 -04:00
  • 645c13592c better py3 support w pre-commit Wing Lian 2023-06-25 10:26:02 -04:00
  • 47d601fa23 optionally define whether to use_fast tokenizer Wing Lian 2023-06-25 10:19:49 -04:00
  • e91fed495a better handling for tokenizers like flan that don't have a bos token flan-no-bos Wing Lian 2023-06-23 15:47:40 -04:00
  • 756dfba97b Merge pull request #218 from OpenAccess-AI-Collective/no-fail-fast Wing Lian 2023-06-23 15:42:54 -04:00
  • 91ab0592af Merge pull request #235 from msinha251/Fixing-data-readme Wing Lian 2023-06-23 13:52:01 -04:00
  • 0aeb7c7802 Fixing Data Readme Mahesh Sinha 2023-06-21 15:34:48 +02:00
  • 9bdd30cdfd Support loading data files from a local directory Utensil 2023-06-21 08:00:58 +00:00
  • d35278aaf1 don't fail fast Wing Lian 2023-06-15 16:01:27 -04:00
  • 9492d4ebb7 Merge pull request #215 from OpenAccess-AI-Collective/adamw-hyperparams-cfg Wing Lian 2023-06-15 12:20:55 -04:00
  • ad5ca4f734 Additional test case per pr Wing Lian 2023-06-15 10:12:47 -04:00
  • cb9d3af5c0 add validation and tests for adamw hyperparam Wing Lian 2023-06-15 09:39:42 -04:00
  • c969f0a9dc add docs Wing Lian 2023-06-15 08:43:20 -04:00
  • 6d0ee4ba34 support adamw and grad norm hyperparams Wing Lian 2023-06-15 08:40:41 -04:00
  • a81f52d575 Merge pull request #212 from OpenAccess-AI-Collective/doc-20230615-v1 Wing Lian 2023-06-15 08:28:57 -04:00
  • 1925eaf1e6 Merge pull request #214 from OpenAccess-AI-Collective/fix-tokenizing-labels Wing Lian 2023-06-15 08:13:43 -04:00
  • 1ab3bf3e67 fix test name Wing Lian 2023-06-15 02:09:33 -04:00
  • d7635b7148 hint to what AMP means Wing Lian 2023-06-15 02:06:27 -04:00
  • 88e17ffc50 add float16 docs and tweak typehints Wing Lian 2023-06-15 00:26:44 -04:00
  • baed440fa1 ingore duplicate code in tests Wing Lian 2023-06-15 02:03:53 -04:00
  • 7925ddce86 bugfix for potential off by one Wing Lian 2023-06-15 01:59:33 -04:00
  • 6f849809c5 Merge pull request #206 from MaciejKarasek/issue205 Wing Lian 2023-06-14 14:23:38 -04:00
  • c16644d05e Merge pull request #209 from sroecker/fix_redpajama_example_tokenizer Wing Lian 2023-06-14 14:23:21 -04:00
  • 945c4191a3 Use AutoTokenizer for redpajama example Steffen Röcker 2023-06-14 20:09:26 +02:00
  • 136522f9c9 style correction maciej.karasek 2023-06-14 20:02:09 +02:00
  • 556fe408b3 issue #205 bugfix maciej.karasek 2023-06-14 16:59:57 +02:00
  • 16bb6276a5 Merge pull request #92 from OpenAccess-AI-Collective/flash-optimum Wing Lian 2023-06-14 07:50:15 -04:00
  • 05d19d2037 remove debugging, use gpt2 since starcoder requires consent no-bos-tokens-packing Wing Lian 2023-06-13 21:32:47 -04:00
  • 61f44f311e fix packing for tokenizers that don't use a bos_token when the bos token and eos token are both the same Wing Lian 2023-06-13 21:26:13 -04:00
  • 06674a11f2 Merge pull request #202 from OpenAccess-AI-Collective/NanoCode012-patch-1 NanoCode012 2023-06-14 09:48:35 +09:00
  • 3513885f43 Fix sharegpt type NanoCode012 2023-06-14 01:10:58 +09:00
  • 06652c1c39 Merge pull request #196 from OpenAccess-AI-Collective/openllama-ft-config v0.2.1 Wing Lian 2023-06-13 11:51:04 -04:00
  • 068fc48978 Merge pull request #199 from NanoCode012/chore/prompter-arg NanoCode012 2023-06-13 17:56:22 +09:00
  • aaadacf6b3 Merge pull request #200 from PocketDocLabs/main Wing Lian 2023-06-13 04:44:34 -04:00
  • 5ff547dc70 Update README.md to include a community showcase PocketDoc Labs 2023-06-12 22:38:10 -07:00
  • dc77c8ebce chore: Refactor inf_kwargs out NanoCode012 2023-06-13 12:01:46 +09:00
  • 51a4c12242 Merge pull request #197 from mhenrichsen/chore/update-readme NanoCode012 2023-06-13 11:53:26 +09:00
  • 4b43a66a0b update alpaca_chat prompts for instructions to explainn the conversation Wing Lian 2023-06-12 18:38:38 -04:00
  • 34ae69989f fix inference mhenrichsen 2023-06-12 21:39:19 +02:00
  • 7dc580b837 add axolotl trainer and quadratic warmup Wing Lian 2023-06-12 00:18:21 -04:00
  • fd2c9814c9 Merge branch 'main' into flash-optimum Wing Lian 2023-06-12 13:12:15 -04:00
  • 2ba4ae8f46 tweak config to work Wing Lian 2023-06-12 10:07:18 -04:00
  • 93dacba228 Merge pull request #187 from OpenAccess-AI-Collective/strip-peft-device-map Wing Lian 2023-06-12 09:10:49 -04:00
  • 8002ffb41f Merge pull request #177 from NanoCode012/fix/landmark-patch Wing Lian 2023-06-12 08:27:12 -04:00
  • 74ef5cc083 Merge pull request #192 from OpenAccess-AI-Collective/sharegpt-custom-prompt Wing Lian 2023-06-12 08:26:38 -04:00
  • 5e616d91c0 Merge branch 'main' into strip-peft-device-map Wing Lian 2023-06-12 08:25:54 -04:00
  • 94f310c7a6 Merge pull request #193 from OpenAccess-AI-Collective/config-fixes-20230612 Wing Lian 2023-06-12 08:24:52 -04:00
  • 8e568bbdae Merge pull request #159 from AngainorDev/patch-1 NanoCode012 2023-06-12 20:27:11 +09:00
  • e21dab49fd Merge pull request #194 from NanoCode012/fix/config-path NanoCode012 2023-06-12 19:28:12 +09:00
  • 52cde69288 Fix config path after config moved NanoCode012 2023-06-12 17:06:15 +09:00
  • 9a58e99e81 config fixes Wing Lian 2023-06-12 01:52:58 -04:00
  • c7dee56b87 add typehints Wing Lian 2023-06-11 19:52:34 -04:00
  • aac4b7691e add new sharegpt, refactor prompt so it can be customized later, add exception if no data is processed Wing Lian 2023-06-11 18:46:26 -04:00