 50h100a
|
273c61d406
guard against nan temperature from dynatemp (or anywhere else).
|
vor 2 Monaten |
|
50h100a
|
a22e887319
why we don't use the github website editor to make changes
|
vor 2 Monaten |
|
50h100a
|
54a8320638
logit shenanigans to prevent even worse shenanigans
|
vor 2 Monaten |
|
50h100a
|
b6a897d2a1
fix temperature, and address those pernicious dynatemp NaNs
|
vor 2 Monaten |
|
50h100a
|
83040c6389
Mask dynatemp using min/max, rather than exp
|
vor 2 Monaten |
 Pyroserenus
|
99fc6f4697
fix: crash in token bans (#764)
|
vor 4 Monaten |
 AlpinDale
|
ad181e3fef
feat: bring back dynatemp (#754)
|
vor 4 Monaten |
|
AlpinDale
|
6329c2d53f
chore: re-enable custom token bans (#751)
|
vor 4 Monaten |
|
AlpinDale
|
abd9d5799a
feat: add XTC Sampling (#740)
|
vor 4 Monaten |
|
AlpinDale
|
198029295c
fix: empty sampler output when temperature is too low (#709)
|
vor 4 Monaten |
|
AlpinDale
|
1394008421
chore: decouple `should_modify_greedy_probs_inplace (#671)
|
vor 4 Monaten |
|
AlpinDale
|
2da6a3ec2b
feat: option to apply temperature scaling last (#670)
|
vor 4 Monaten |
 50h100a
|
cbdf2d986f
quadratic sampling: separate diff from logits to avoid NaNs.
|
vor 5 Monaten |
|
AlpinDale
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
vor 5 Monaten |
|
AlpinDale
|
9d81716bfd
[v0.5.3] Release Candidate (#388)
|
vor 8 Monaten |
|
50h100a
|
f663d3fccc
Merge pull request #397 from 50h100a/pr_samplerasserts
|
vor 10 Monaten |
|
50h100a
|
85ae23ac3c
Missed .items() and assert
|
vor 10 Monaten |
|
50h100a
|
43c9858854
Merge pull request #244 from PygmalionAI/faster_topk
|
vor 10 Monaten |
|
50h100a
|
bd564148e2
Merge branch 'main' of https://github.com/PygmalionAI/aphrodite-engine into ffs
|
vor 10 Monaten |
|
50h100a
|
d3dd170a7d
merge main
|
vor 10 Monaten |
|
AlpinDale
|
78d66f16d1
Chunked Prefill Part 1 (#384)
|
vor 10 Monaten |
|
AlpinDale
|
9181fa0396
feat: Triton kernels for sampling (#383)
|
vor 10 Monaten |
|
50h100a
|
dc09dc2b4d
Merge branch 'main' into pr_samplers
|
vor 10 Monaten |
|
AlpinDale
|
f8dfac6372
chore: attention refactor and upstream sync apr01 (#365)
|
vor 10 Monaten |
|
50h100a
|
7ed57e318d
Overhauled SamplingTensors construction.
|
vor 10 Monaten |
|
50h100a
|
d5dbd29db4
hoist sampler internals into a single function.
|
vor 10 Monaten |
|
AlpinDale
|
da223153c6
feat&fix: cohere support and missing GPU blocks (#333)
|
vor 10 Monaten |
|
AlpinDale
|
e42a78381a
feat: switch from pylint to ruff (#322)
|
vor 10 Monaten |
|
AlpinDale
|
9fa99215f8
feat: add cubic sampling (#280)
|
vor 11 Monaten |
|
AlpinDale
|
657aec0cbd
refactor: OpenAI endpoint (#261)
|
vor 11 Monaten |
