- 29 Apr, 2022 4 commits
- 28 Apr, 2022 5 commits
-
-
Tian Yun authored
-
Tian Yun authored
-
Tian Yun authored
-
tomlimi authored
-
Zdenek Kasner authored
-
- 27 Apr, 2022 8 commits
-
-
jon-tow authored
-
Tian Yun authored
-
cjlovering authored
-
cjlovering authored
-
cjlovering authored
-
Zdenek Kasner authored
-
cjlovering authored
- saves the doc idx by order in dataset - saves the full ctx. we can turn this off it is too much memory - saves the idx of the fewshot examples - saves the source (test/val) of the fewshot examples - TODO: We now need to return the pred/target from process_results - TODO: Make it so override-ed process-resutls don't have to re-impl this.
-
jon-tow authored
-
- 26 Apr, 2022 11 commits
-
-
cjlovering authored
-
cjlovering authored
Updated doc: If the answer choices is empty, then it is generation; else ranked choice. This will be the canonical approach when using PS.
-
cjlovering authored
-
cjlovering authored
-
cjlovering authored
-
cjlovering authored
-
cjlovering authored
-
cjlovering authored
-
jon-tow authored
Turned off generation tasks for now. Changed process to look at the metrics. Only accuracy implemented.
-
jon-tow authored
-
jon-tow authored
-
- 25 Apr, 2022 3 commits
-
-
jon-tow authored
-
jon-tow authored
-
cjlovering authored
-
- 31 Mar, 2022 3 commits
- 18 Mar, 2022 1 commit
-
-
Jonathan Tow authored
-
- 04 Jan, 2022 1 commit
-
-
Jonathan Tow authored
-
- 25 Dec, 2021 1 commit
-
-
Igor Ostrovsky authored
bits_per_byte was calculated as average per-byte loglikelihood, which would work if loglikelihood was base-2 log, but it is natural log. To correct for that, bits_per_byte should be divided by math.log(2). Also, it should be true that 2^bits_per_byte == byte_perplexity, which is true after the fix.
-
- 24 Dec, 2021 1 commit
-
-
Jonathan Tow authored
-
- 21 Dec, 2021 1 commit
-
-
Jonathan Tow authored
-
- 16 Dec, 2021 1 commit
-
-
Jonathan Tow authored
-