 Max Ryabinin
|
4159e557bf
Create dummy data when materializing qkv_proj
|
пре 1 година |
|
Max Ryabinin
|
9cb4c721e7
Fix checking for nonexistent keys
|
пре 1 година |
|
Max Ryabinin
|
16fb547960
Fix removal of nonexistent keys
|
пре 1 година |
|
Max Ryabinin
|
f100915641
Reformat code with black
|
пре 1 година |
|
Max Ryabinin
|
4644131086
Ignore missing qkv_proj.weight when loading a checkpoint
|
пре 1 година |
|
Max Ryabinin
|
b2ab84cc33
Add dry_run option to --throughput
|
пре 1 година |
|
Max Ryabinin
|
c666a975d0
Remove unused import in throughput.py
|
пре 1 година |
|
Max Ryabinin
|
57119bb201
Merge query/key/value projection layers
|
пре 1 година |
 Alexander Borzunov
|
b4d822afb2
Force use_cache=True in config only (#497)
|
пре 1 година |
|
Alexander Borzunov
|
abd547735f
Force use_cache=True (#496)
|
пре 1 година |
|
Alexander Borzunov
|
6ef6bf5fa2
Create model index in DHT (#491)
|
пре 2 година |
|
Alexander Borzunov
|
6bb3f54e39
Replace dots in repo names when building DHT prefixes (#489)
|
пре 2 година |
|
Alexander Borzunov
|
02fc71eb25
Fix race condition in MemoryCache (#487)
|
пре 2 година |
|
Alexander Borzunov
|
dc0072fde1
Wait for DHT storing state OFFLINE on shutdown (#486)
|
пре 2 година |
|
Alexander Borzunov
|
a26559ff65
Fix `.generate(input_ids=...)` (#485)
|
пре 2 година |
|
Alexander Borzunov
|
459933f846
Remove no-op process in PrioritizedTaskPool (#484)
|
пре 2 година |
|
Alexander Borzunov
|
26ebbfe8f0
Support macOS (#477)
|
пре 2 година |
|
Alexander Borzunov
|
75e516a8c1
Refactor readme (#482)
|
пре 2 година |
 justheuristic
|
c08d09c4d3
Rewrite MemoryCache alloc_timeout logic (#434)
|
пре 2 година |
|
Alexander Borzunov
|
90840dfea2
Fix requiring transformers>=4.32.0 (#480)
|
пре 2 година |
|
Alexander Borzunov
|
915b357740
Require transformers>=4.32.0 (#479)
|
пре 2 година |
|
Alexander Borzunov
|
18e93afc73
Don't install cpufeature on non-x86_64 machines (#478)
|
пре 2 година |
|
Alexander Borzunov
|
6967904590
Bump version to 2.1.0 (#474)
|
пре 2 година |
|
Alexander Borzunov
|
df8ab09ca2
Hide excess key message (#476)
|
пре 2 година |
 Artem Chumachenko
|
a14ae7334d
Update peft to 0.5.0 version (#475)
|
пре 2 година |
|
Alexander Borzunov
|
a9b0e9ff1a
Support loading weights from Safetensors on server (#473)
|
пре 2 година |
|
justheuristic
|
4f850996bb
Change transformers version assert (#472)
|
пре 2 година |
|
justheuristic
|
9250025140
Support transformers 4.32.x (#471)
|
пре 2 година |
|
justheuristic
|
adda5f8c20
Temporarily require peft<0.5.0, transformers<4.32.0 (#470)
|
пре 2 година |
|
Alexander Borzunov
|
de2475f31c
Make client compatible with transformers' GenerationMixin (#464)
|
пре 2 година |
