aLoRA Support #15327
gabe-l-hart
merged
19 commits into
ggml-org:master
from
gabe-l-hart:gabe-l-hart/alora-support
feat: Add python-side constants and conversion for adapter.lora.invoc…
5d42ca60
feat: Add c++ side constants for adapter.lora.invocation_string
a242ee8c
feat: Parse invocation string for adapters from GGUF
4c214e42
fix(python): Update conversion to alora_invocation_tokens
636bb2b0
fix(cpp): Update to alora_invocation_tokens on c++ side
215841fc
feat: Add C APIs to get alora invocation token array from lora
7212b9de
feat: Initial implementation of alora cache logic in server
45eab0df
feat: Identify alora invocation sequences
457c757e
feat: Only reuse cache for tokens before the alora invocation start
a5a75da1
feat: Handle un-cached tokens that come before the alora activation
b4a0cecc
fix: Use || instead of 'or'
d03d1067
fix: Fix off-by-one for limiting cached tokens to before alora start
52b37e13
fix: Support backwards-compatibility for "invocation_string" in adapt…
da2e8c1e
Merge remote-tracking branch 'origin/master' into gabe-l-hart/alora-s…
37d0bd0b
Merge remote-tracking branch 'origin/master' into gabe-l-hart/alora-s…
f863c9f7
Merge remote-tracking branch 'origin/master' into gabe-l-hart/alora-s…
d8651c8e
CISC
approved these changes
on 2025-09-05
fix: Remove duplicate logging
5958557f
Merge remote-tracking branch 'origin/master' into gabe-l-hart/alora-s…
a29364dc
feat: Report alora_invocation_string and alora_invocation_tokens from…
56476765
Login to write a write a comment.
Login via GitHub