What's Changed
- Default settings adjustment by @psinger in #154
- Manual loss & chained setting by @psinger in #142
- support azure endpoints for openai api by @pascal-pfeiffer in #147
- update transformers to 4.30.1 by @maxjeblick in #158
- fix for additional pretraining and more detailed FAQ by @pascal-pfeiffer in #155
- switch back to default chat behaviour by @pascal-pfeiffer in #159
- enable train progress bar logging for sub 1 epoch evals by @pascal-pfeiffer in #162
- added perplexity as metric by @pascal-pfeiffer in #157
- personalize chatbot by @pascal-pfeiffer in #161
- hotfix by @pascal-pfeiffer in #164
- fix for custom chatbot_name and chatbot_author by @pascal-pfeiffer in #165
- Add documentation about updating H2O LLM Studio by @maxjeblick in #169
- Pp/rlhf by @pascal-pfeiffer in #152
- Pp/typing by @pascal-pfeiffer in #173
- Allow to change branch for downloading huggingface models by @haqishen in #168
- Use huggingface auth token for downloading models by @maxjeblick in #167
- Dataset refactoring by @maxjeblick in #177
- fix style command by @maxjeblick in #178
- Fix auth token for falcon models by @maxjeblick in #181
- Package update by @psinger in #176
- Run llmstudio docker image as llmstudio user instead of root user by @tomkraljevic in #163
- Dtype & HF Push Changes by @psinger in #182
- Use cache changes by @psinger in #188
- Add falcon peft target modules by @maxjeblick in #166
- Def loss in dictionary by @psinger in #193
- GPT metric endpoint by @psinger in #195
- Refactoring by @maxjeblick in #192
- Make gpu id for chat configurable by @maxjeblick in #184
- Update wave by @maxjeblick in #198
- Update README.md by @psinger in #199
- Push official docs site for LLM Studio by @sherenem in #156
- Fix documentation base URL by @sherenem in #201
- Diverse adjustments by @psinger in #207
- Fix link to documentation in README by @tmm1 in #202
- rlhf batches by @pascal-pfeiffer in #197
- Add doc issue template by @sherenem in #204
- check available disk space by @haqishen in #194
- Wave streaming by @maxjeblick in #205
- Checkpoint changes by @psinger in #222
- add torch uint8 for disk check by @haqishen in #224
- Fix tokenizer length by @psinger in #231
- Fix chat bug by @maxjeblick in #230
- cast to float before converting to numpy by @pascal-pfeiffer in #234
- make lora slider max values adjustable by @pascal-pfeiffer in #236
- Bitsandbytes fixes by @psinger in #226
- version bump to 0.0.4 and default batch_size 2 by @pascal-pfeiffer in #225
New Contributors
- @haqishen made their first contribution in #168
- @tomkraljevic made their first contribution in #163
- @tmm1 made their first contribution in #202
Full Changelog: v0.0.3...v0.0.4