 Blake Mizerany
|
cb42e607c5
llm: speed up gguf decoding by a lot (#5246)
|
10 mēneši atpakaļ |
 Daniel Hiltgen
|
6f351bf586
review comments and coverage
|
11 mēneši atpakaļ |
|
Daniel Hiltgen
|
ff4f0cbd1d
Prevent multiple concurrent loads on the same gpus
|
11 mēneši atpakaļ |
|
Daniel Hiltgen
|
fc37c192ae
Refine CPU load behavior with system memory visibility
|
11 mēneši atpakaļ |
|
Daniel Hiltgen
|
434dfe30c5
Reintroduce nvidia nvml library for windows
|
11 mēneši atpakaļ |
|
Daniel Hiltgen
|
48702dd149
Harden unload for empty runners
|
11 mēneši atpakaļ |
|
Daniel Hiltgen
|
5e8ff556cb
Support forced spreading for multi GPU
|
1 gadu atpakaļ |
 Michael Yang
|
e40145a39d
lint
|
11 mēneši atpakaļ |
|
Michael Yang
|
c895a7d13f
some gocritic
|
11 mēneši atpakaļ |
|
Michael Yang
|
04f3c12bb7
replace x/exp/slices with slices
|
11 mēneši atpakaļ |
 Patrick Devine
|
4cc3be3035
Move envconfig and consolidate env vars (#4608)
|
11 mēneši atpakaļ |
 Sang Park
|
4434d7f447
Correct typo in error message (#4535)
|
11 mēneši atpakaļ |
|
Daniel Hiltgen
|
ec231a7923
Remove VRAM convergence check for windows
|
11 mēneši atpakaļ |
|
Patrick Devine
|
6845988807
Ollama `ps` command for showing currently loaded models (#4327)
|
11 mēneši atpakaļ |
|
Daniel Hiltgen
|
4142c3ef7c
Always use the sorted list of GPUs
|
11 mēneši atpakaļ |
 Jeffrey Morgan
|
bb6fd02298
Don't clamp ctx size in `PredictServerFit` (#4317)
|
11 mēneši atpakaļ |
|
Daniel Hiltgen
|
354ad9254e
Wait for GPU free memory reporting to converge
|
11 mēneši atpakaļ |
|
Jeffrey Morgan
|
c9f98622b1
Skip scheduling cancelled requests, always reload unloaded runners (#4189)
|
1 gadu atpakaļ |
|
Jeffrey Morgan
|
dfa2f32ca0
unload in critical section (#4187)
|
1 gadu atpakaļ |
|
Daniel Hiltgen
|
f56aa20014
Centralize server config handling
|
1 gadu atpakaļ |
|
Jeffrey Morgan
|
942c979232
allocate a large enough kv cache for all parallel requests (#4162)
|
1 gadu atpakaļ |
|
Daniel Hiltgen
|
20f6c06569
Make maximum pending request configurable
|
1 gadu atpakaļ |
 Mark Ward
|
63c763685f
log when the waiting for the process to stop to help debug when other tasks execute during this wait.
|
1 gadu atpakaļ |
|
Mark Ward
|
f4a73d57a4
fix runner expire during active use. Clearing the expire timer as it is used. Allowing the finish to assign an expire timer so that the runner will expire after no use.
|
1 gadu atpakaļ |
|
Daniel Hiltgen
|
d6e3b64582
Fix concurrency for CPU mode
|
1 gadu atpakaļ |
 Daniel Hiltgen
|
9b5a3c5991
Merge pull request #3914 from dhiltgen/mac_perf
|
1 gadu atpakaļ |
|
Jeffrey Morgan
|
00b0699c75
Reload model if `num_gpu` changes (#3920)
|
1 gadu atpakaļ |
|
Daniel Hiltgen
|
b123be5b71
Adjust context size for parallelism
|
1 gadu atpakaļ |
 Bryce Reitano
|
36a6daccab
Restructure loading conditional chain
|
1 gadu atpakaļ |
|
Bryce Reitano
|
284e02bed0
Move ggml loading to when we attempt fitting
|
1 gadu atpakaļ |
