Commits · b49dbf2d88c340c3686e4318985d8b64581364b7 · OpenDAS / text-generation-inference

16 Mar, 2023 1 commit
- fix(server): use server tokenizer as gt (#128) · b49dbf2d
  OlivierDehaene authored Mar 16, 2023
  
  b49dbf2d
15 Mar, 2023 1 commit
- fix(server): add position ids to neox (#126) · 8ad60b75
  OlivierDehaene authored Mar 15, 2023
  
  8ad60b75
13 Mar, 2023 1 commit
- fix(server): revert gpt-neox optims (#123) · cbd36aa4
  OlivierDehaene authored Mar 13, 2023
  
  cbd36aa4
09 Mar, 2023 3 commits
- v0.4.0 (#119) · 411d6247
  OlivierDehaene authored Mar 09, 2023
  
  411d6247
- fix(server): do not warp prefill logits (#116) · c0795de2
  OlivierDehaene authored Mar 09, 2023
  
  c0795de2
- feat: support typical sampling (#114) · 1a2d6825
  OlivierDehaene authored Mar 09, 2023
```
closes #112
```
  1a2d6825
08 Mar, 2023 1 commit
- fix(server): fix index out of range for watermarking (#110) · 941cd42e
  OlivierDehaene authored Mar 08, 2023
  
  941cd42e
07 Mar, 2023 2 commits
- fix(server): fix galactica batch (#106) · b1485e18
  OlivierDehaene authored Mar 07, 2023
```
closes #105
```
  b1485e18
- feat(clients): Python client (#103) · 3fef90d5
  OlivierDehaene authored Mar 07, 2023
  
  3fef90d5
06 Mar, 2023 2 commits
- feat: allow local models (#101) · cd5961b5
  OlivierDehaene authored Mar 06, 2023
```
closes #99
```
  cd5961b5
- fix(server): fix generate_stream by forcing tokens to be decoded correctly (#100) · 9b205d33
  OlivierDehaene authored Mar 06, 2023
  
  9b205d33
03 Mar, 2023 3 commits
- v0.3.2 (#97) · 1c19b093
  OlivierDehaene authored Mar 03, 2023
  
  1c19b093
- feat(server): fix transformers commit (#96) · 0b6807ca
  OlivierDehaene authored Mar 03, 2023
  
  0b6807ca
- feat(server): update to hf_transfer==0.1.2 (#93) · 2d39f199
  OlivierDehaene authored Mar 03, 2023
  
  2d39f199
02 Mar, 2023 1 commit
- feat(server): add logits watermark (#90) · 9b8ea6a6
  OlivierDehaene authored Mar 02, 2023
  
  9b8ea6a6
24 Feb, 2023 4 commits
- fix(server): fix token_is_special (#87) · 65e2f162
  OlivierDehaene authored Feb 24, 2023
  
  65e2f162
- feat(server): add special token bool (#85) · 0ac184ce
  OlivierDehaene authored Feb 24, 2023
  
  0ac184ce
- v0.3.1 (#84) · 4b1c9720
  OlivierDehaene authored Feb 24, 2023
  
  4b1c9720
- feat(server): pre-allocate max attention mask (#75) · 44ce098c
  OlivierDehaene authored Feb 24, 2023
  
  44ce098c
20 Feb, 2023 1 commit
- fix(server): remove position_ids from galactica forward (#82) · 78063c05
  OlivierDehaene authored Feb 20, 2023
```
closes #80
```
  78063c05
18 Feb, 2023 1 commit
- feat(server): enable hf-transfer (#76) · 17bc841b
  OlivierDehaene authored Feb 18, 2023
  
  17bc841b
16 Feb, 2023 2 commits
- v0.3.0 (#72) · c720555a
  OlivierDehaene authored Feb 16, 2023
  
  c720555a
- feat(router): add prometheus metrics scrape endpoint (#71) · 439fcaf8
  OlivierDehaene authored Feb 16, 2023
  
  439fcaf8
15 Feb, 2023 1 commit
- feat(server): improve download logging (#66) · c5a4a1fa
  OlivierDehaene authored Feb 15, 2023
  
  c5a4a1fa
14 Feb, 2023 1 commit
- feat: add safetensors conversion (#63) · 0fbc6919
  OlivierDehaene authored Feb 14, 2023
  
  0fbc6919
13 Feb, 2023 1 commit
- feat: add distributed tracing (#62) · 9af45414
  OlivierDehaene authored Feb 13, 2023
  
  9af45414
08 Feb, 2023 1 commit
- fix(docker): increase shm size (#60) · 1ad3250b
  OlivierDehaene authored Feb 08, 2023
  
  1ad3250b
07 Feb, 2023 3 commits
- feat(server): support t5 (#59) · c503a639
  OlivierDehaene authored Feb 07, 2023
  
  c503a639
- V0.2.1 (#58) · 2fe5e1b3
  OlivierDehaene authored Feb 07, 2023
  
  2fe5e1b3
- fix(server): better handling of inference mode (#57) · 4acc42a6
  OlivierDehaene authored Feb 07, 2023
  
  4acc42a6
03 Feb, 2023 1 commit
- feat(router): refactor API and add openAPI schemas (#53) · 20c3c594
  OlivierDehaene authored Feb 03, 2023
  
  20c3c594
02 Feb, 2023 2 commits
- breaking(router): modify /generate API to only return generated text (#50) · b1482d90
  OlivierDehaene authored Feb 02, 2023
```
@njhill, @yk FYI

generated_text was concatenated to the user prompt for legacy reason. We
want to remove this behaviour as we don't think it is useful and even
detrimonial to usability.

We also remove the unused Vec.
```
  b1482d90
- fix(server): allow greedy repetition penalty (#51) · df227ac2
  OlivierDehaene authored Feb 02, 2023
  
  df227ac2
01 Feb, 2023 3 commits
- feat(server): allow the server to use a local weight cache (#49) · 775115e3
  OlivierDehaene authored Feb 01, 2023
  
  775115e3
- feat(server): support repetition penalty (#47) · 313194f6
  OlivierDehaene authored Feb 01, 2023
  
  313194f6
- feat(server): allow gpt-neox models with odd vocab sizes to be sharded (#48) · 2ad895a6
  OlivierDehaene authored Feb 01, 2023
  
  2ad895a6
31 Jan, 2023 4 commits
- feat(server): Support GPT-Neox (#39) · f830706b
  OlivierDehaene authored Jan 31, 2023
  
  f830706b
- fix(server): fix quantization for sharded models (#45) · c6e8b944
  OlivierDehaene authored Jan 31, 2023
  
  c6e8b944
- feat: Add token streaming using ServerSideEvents support (#41) · 017a2a8c
  OlivierDehaene authored Jan 31, 2023
  
  017a2a8c
- fix(server): fix seeding with multiple shards (#44) · 54fec931
  OlivierDehaene authored Jan 31, 2023
  
  54fec931