 AlpinDale
|
d8a51d05a7
fix: seeded gens with pipeline parallel
|
5 月之前 |
|
AlpinDale
|
2c653a2268
fix: make speculative decoding work with per-request seed
|
5 月之前 |
 AlpinDale
|
af43576da0
feat: add MLPSpeculator speculative decoding support (#572)
|
5 月之前 |
|
AlpinDale
|
4d1e613804
chore: minor simplifications
|
5 月之前 |
|
AlpinDale
|
e0886ee929
feat: add `ProposerWorkerBase` abstract class
|
6 月之前 |
|
AlpinDale
|
a94de94c44
refactor: combine the prefill and decode into a single API (#553)
|
6 月之前 |
|
AlpinDale
|
ef733aee43
implement ExecuteModelData to reduce executor complexity
|
6 月之前 |
|
AlpinDale
|
79901b76de
logprobs for target model (spec decoding)
|
6 月之前 |
|
AlpinDale
|
723c6acb84
re-add ngram speculative decoding
|
6 月之前 |
|
AlpinDale
|
9d81716bfd
[v0.5.3] Release Candidate (#388)
|
8 月之前 |
|
AlpinDale
|
f8dfac6372
chore: attention refactor and upstream sync apr01 (#365)
|
10 月之前 |
