What's Changed
- added markdown tooltip by @Anantha-Kandrapu in #2064
- duplex communication in inference by @yk in #2098
- Fix passing parallelism var to worker in Ansible playbook by @olliestanley in #2092
- Add XP progress by @AbdBarho in #2093
- Update Chat UI to latest inference changes by @AbdBarho in #2105
- Inference-deploy by @yk in #2109
- Inference-comma-separated-list-configs by @yk in #2110
- added create_at and parent_id to inference MessageRead by @yk in #2111
- Improve FAQ by @olliestanley in #2114
- Add GitHub OAuth support to inference server by @olliestanley in #2097
- Clean up chat UI by @AbdBarho in #2108
- Add Team page by @AbdBarho in #2103
- Added James Mete (JMete) to team page. by @Jmete in #2117
- Update team.json by @melvinebenezer in #2116
- Minor update on team.json by @melvinebenezer in #2118
- Some changes in teams by @0x22almostEvil in #2120
- Reward Model Training by @dvruette in #2071
- You in teams and more by @0x22almostEvil in #2123
- Add minimath support by @johnflux in #2094
- Add
Zyto /website/src/data/team.json by @Zyphre in #2125 - Fix GitHub auth flow on inference server by @olliestanley in #2121
- Add tlcv2.0_oa by @wannaphong in #1965
- Basic-hf-server by @yk in #2130
- feature: Using Celery for APIs and periodic tasks by @melvinebenezer in #1754
- enable chat via environment variable instead of flag by @yk in #2128
- Boilerplate OAuth website <=> inference by @AbdBarho in #2127
- Add Joak to team by @notmd in #2134
- Evaluation datasets for Reward Model by @shahules786 in #2112
- fix team description by @notmd in #2135
- Add me to team.json by @johnflux in #2132
- Update team.json by @kenhktsui in #2139
- Simplify celery worker use by @andreaskoepf in #2131
- added HF, W&B, and Redmond to sponsors by @yk in #2142
- Display message tree state admin by @Rulios in #2133
- Add multi-turn conversation dataset format by @Vechtomov in #2113
- [#2077] Possbility to revert message deletion by @Shadowner in #2079
- Added ENABLE_CHAT to chat id page and docker-compose by @yk in #2137
- Llama inference by @yk in #2144
- hotfix for max_new_tokens parameter being None by @yk in #2146
- Implemented ability to press ctrl+enter (or cmd+enter) to Review/Subm… by @cooperaustinj in #1237
- Cancelling pending chat messages of user on submission of a new message by @yk in #2141
- Added a timeout for inference messages by @yk in #2147
- Added various people to codeowners for backend and inference by @yk in #2150
- Remove invalid badge from README by @olliestanley in #2149
- Add refresh tokens to inference server auth by @olliestanley in #2129
- Robustification: Specific message type for protocol upgrade, everything else will re-connect by @yk in #2152
- Fixes auth TokenPair change bug (also fix for null logprob from inference server) by @yk in #2155
- Fixes changed access token model for website. Does not handle refresh yet by @yk in #2157
- Updated team page to fit visually more the the rest of the pages, added a button to the call-to-action section which links to the team page. by @theopfr in #2126
- Fix stacking of pos/neg logits for RM compute_metrics by @andreaskoepf in #2138
- building separate llama image for inference by @yk in #2161
- Changed worker config into worker info and unified model input and total lengths by @yk in #2160
- Added hardware info dependencies to oasst-shared to make api tests pass by @yk in #2162
- Derive RewardModel from PreTrainedModel by @andreaskoepf in #2158
- check
per_digit_tokens
exists by @HuangLK in #2166 - Only allow prompter answers to complete parent message by @yk in #2164
- Add first RM training run params by @andreaskoepf in #2167
- Diabling stop sequences in inference, since we are using eos_tokens in most models by @yk in #2168
- Added param configs from eval to pass to the frontend by @yk in #2163
- Fix: Correctly sending error to waiting client on message abort by @yk in #2169
- Chat config UI by @notmd in #2115
- Update pythia 6b RM training config by @andreaskoepf in #2187
- Add @dvruette, @andreaskoepf & @yk as codeowners of /model/ by @andreaskoepf in #2186
- Fix inconsistent sources of max token length by @yk in #2178
- Add GPTNeoXRewardModel by @andreaskoepf in #2182
- Add llama training support by @andreaskoepf in #2055
- Fix output of RankingDataCollator when max_length is None by @andreaskoepf in #2185
- Update website's dockerfile by @AbdBarho in #2189
- Login to inference through website by @AbdBarho in #2176
- rework layout by @notmd in #2172
- Evaluate sampling report using RM by @shahules786 in #2190
New Contributors
- @Anantha-Kandrapu made their first contribution in #2064
- @Zyphre made their first contribution in #2125
- @Rulios made their first contribution in #2133
- @cooperaustinj made their first contribution in #1237
- @HuangLK made their first contribution in #2166
Full Changelog: v0.0.1-beta56...v0.0.1-beta57