Commits · 0bb597b42c53355a567aba2a1357cc34b9d99ddd · OpenDAS / Megatron-LM

18 Jul, 2022 1 commit
- Adding checks for total number of tokens to keep server from crashing · cfe2c2be
  rprenger authored Jul 18, 2022
  
  cfe2c2be
21 Jun, 2022 1 commit
- Changing name of parameters from factual_decay to top_p_decay · ed6806ac
  rprenger authored Jun 21, 2022
  
  ed6806ac
08 Jun, 2022 1 commit
- Adding top_p decay and bound for factual sampling from Factuality Enhanced LMs · b49349ec
  rprenger authored Jun 08, 2022
  
  b49349ec
16 May, 2022 1 commit
- Fixing beam search in distributed mode · 6be75e2a
  rprenger authored May 16, 2022
  
  6be75e2a
05 May, 2022 1 commit
- Making the API talk to the server, fixed a bug where parameters weren't getting forwarded to node · fd176a90
  rprenger authored May 04, 2022
  
  fd176a90
19 Feb, 2022 2 commits
- Possibly working beam search · de84b2af
  rprenger authored Feb 18, 2022
  
  de84b2af
- Almost working beam search · fa75238b
  rprenger authored Feb 18, 2022
  
  fa75238b
27 Jan, 2022 1 commit
- Adding the option to not log · 31fd62d6
  rprenger authored Jan 26, 2022
  
  31fd62d6
02 Nov, 2021 2 commits
- Removing old comment · acaabedd
  rprenger authored Nov 02, 2021
  
  acaabedd
- Adding stop token logic and random seed for deterministic answers · 9b131fad
  rprenger authored Nov 02, 2021
  
  9b131fad
27 Oct, 2021 1 commit
- Remove IP block · 83bc79d6
  rprenger authored Oct 26, 2021
  
  83bc79d6
22 Oct, 2021 1 commit
- Hacked in way to to have stop tokens · 9e0ee6fd
  rprenger authored Oct 22, 2021
  
  9e0ee6fd
20 Oct, 2021 1 commit
- getting rid of unnecessary just_score · 5ab28fd1
  rprenger authored Oct 20, 2021
  
  5ab28fd1
19 Oct, 2021 1 commit
- Code that keeps it from dying when the input prompts are too long · 554bb262
  rprenger authored Oct 19, 2021
  
  554bb262
18 Oct, 2021 1 commit
- Fixing logic in server accepting tokens_to_generate=0 · a3770921
  rprenger authored Oct 18, 2021
  
  a3770921
16 Oct, 2021 1 commit
- Getting tokens_to_generate=0 to work · 9cc286ba
  rprenger authored Oct 15, 2021
  
  9cc286ba
15 Oct, 2021 3 commits
- Allowing for a 0 tokens/just scoring mode · 21d2b0fe
  rprenger authored Oct 15, 2021
  
  21d2b0fe
- removed greedy argument · 71359e1f
  mshoeybi authored Oct 15, 2021
  
  71359e1f
- removed return all probs · c6e7c7fd
  mshoeybi authored Oct 15, 2021
  
  c6e7c7fd
14 Oct, 2021 1 commit
- Changing defaults and query sanitation to keep it from crashing on reasonable queries · 8d405805
  rprenger authored Oct 14, 2021
  
  8d405805
13 Oct, 2021 1 commit
- changed inference to text generation · ee536123
  mshoeybi authored Oct 13, 2021
  
  ee536123
11 Oct, 2021 1 commit
- added new inference to the server · a5bfc296
  mshoeybi authored Oct 10, 2021
  
  a5bfc296
07 Oct, 2021 1 commit
- Forgot to have the outputs named text instead of prompts · 397714f5
  rprenger authored Oct 07, 2021
  
  397714f5
02 Oct, 2021 1 commit
- Adding top_p, top_k, logprobs, prompts, and api endpoint to API · 4b11fd9d
  rprenger authored Oct 01, 2021
  
  4b11fd9d
23 Sep, 2021 2 commits
- Had bugs in the fix that I didn't notice until using the big server · 02c00ce6
  rprenger authored Sep 23, 2021
  
  02c00ce6
- switching from semaphore to lock · 7b293d9b
  rprenger authored Sep 22, 2021
  
  7b293d9b
21 Sep, 2021 1 commit
- addressing comments · f65a0f88
  rprenger authored Sep 21, 2021
  
  f65a0f88
20 Sep, 2021 2 commits
- Fixing the URL for the web interface · 7bdeb1e7
  rprenger authored Sep 20, 2021
  
  7bdeb1e7
- Adding the option for beginning of sentence token (and fixing hangs) · 69757f9a
  rprenger authored Sep 20, 2021
  
  69757f9a
17 Sep, 2021 1 commit

Add temperature to the server API · 527e07c0

Robert Clark authored Sep 10, 2021



A temperature value between 0.0 and 100.0 can now be specified via the API
while running the text generation server. The value passed to
--temperature while running the text generation server is kept as the
default value for all API calls that don't include temperature, even if
different values were manually specified previously.
Signed-Off-By: Robert Clark <roclark@nvidia.com>

527e07c0

14 Sep, 2021 2 commits
- Fixes a bug in broadcasting that was causing hanging · b46482e8
  rprenger authored Sep 14, 2021
  
  b46482e8
- Changing api to tokens_to_generate, making it so we always generate at least tokens_to_generate · 77979e3b
  rprenger authored Sep 14, 2021
  
  77979e3b
27 Aug, 2021 1 commit
- Adding API server · 3fe6821a
  Ryan Prenger authored Aug 27, 2021
  
  3fe6821a
26 Aug, 2021 1 commit
- Addressing comments · 055a673e
  rprenger authored Aug 25, 2021
  
  055a673e
12 Aug, 2021 1 commit
- Got it working on the full big model · 811183f0
  rprenger authored Aug 11, 2021
  
  811183f0
11 Aug, 2021 1 commit
- Got the probs piped · ddd36145
  rprenger authored Aug 11, 2021
  
  ddd36145
29 Jul, 2021 1 commit
- Outputting log probabilities · e0bf5199
  rprenger authored Jul 29, 2021
  
  e0bf5199
20 Jul, 2021 1 commit
- Made some simplifications that keep it from hanging · 59414b33
  rprenger authored Jul 20, 2021
  
  59414b33
19 Jul, 2021 1 commit
- Removing bug possibilities and adding timing info · 7a9c4a03
  rprenger authored Jul 19, 2021
  
  7a9c4a03
09 Jul, 2021 1 commit
- Refactoring code so server code is more independent of sampling and adding a... · 29dd0a35
  rprenger authored Jul 09, 2021
```
Refactoring code so server code is more independent of sampling and adding a CLI.  CLI still has URL of server hard-coded
```
  29dd0a35