 Blake Mizerany
|
cb42e607c5
llm: speed up gguf decoding by a lot (#5246)
|
10 months ago |
 Daniel Hiltgen
|
6f351bf586
review comments and coverage
|
11 months ago |
|
Daniel Hiltgen
|
ff4f0cbd1d
Prevent multiple concurrent loads on the same gpus
|
11 months ago |
|
Daniel Hiltgen
|
fc37c192ae
Refine CPU load behavior with system memory visibility
|
11 months ago |
|
Daniel Hiltgen
|
434dfe30c5
Reintroduce nvidia nvml library for windows
|
11 months ago |
|
Daniel Hiltgen
|
48702dd149
Harden unload for empty runners
|
11 months ago |
|
Daniel Hiltgen
|
5e8ff556cb
Support forced spreading for multi GPU
|
11 months ago |
 Michael Yang
|
e40145a39d
lint
|
11 months ago |
|
Michael Yang
|
c895a7d13f
some gocritic
|
11 months ago |
|
Michael Yang
|
04f3c12bb7
replace x/exp/slices with slices
|
11 months ago |
 Patrick Devine
|
4cc3be3035
Move envconfig and consolidate env vars (#4608)
|
11 months ago |
 Sang Park
|
4434d7f447
Correct typo in error message (#4535)
|
11 months ago |
|
Daniel Hiltgen
|
ec231a7923
Remove VRAM convergence check for windows
|
11 months ago |
|
Patrick Devine
|
6845988807
Ollama `ps` command for showing currently loaded models (#4327)
|
11 months ago |
|
Daniel Hiltgen
|
4142c3ef7c
Always use the sorted list of GPUs
|
11 months ago |
 Jeffrey Morgan
|
bb6fd02298
Don't clamp ctx size in `PredictServerFit` (#4317)
|
11 months ago |
|
Daniel Hiltgen
|
354ad9254e
Wait for GPU free memory reporting to converge
|
11 months ago |
|
Jeffrey Morgan
|
c9f98622b1
Skip scheduling cancelled requests, always reload unloaded runners (#4189)
|
1 year ago |
|
Jeffrey Morgan
|
dfa2f32ca0
unload in critical section (#4187)
|
1 year ago |
|
Daniel Hiltgen
|
f56aa20014
Centralize server config handling
|
1 year ago |
|
Jeffrey Morgan
|
942c979232
allocate a large enough kv cache for all parallel requests (#4162)
|
1 year ago |
|
Daniel Hiltgen
|
20f6c06569
Make maximum pending request configurable
|
1 year ago |
 Mark Ward
|
63c763685f
log when the waiting for the process to stop to help debug when other tasks execute during this wait.
|
1 year ago |
|
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 year ago |
|
Daniel Hiltgen
|
d6e3b64582
Fix concurrency for CPU mode
|
1 year ago |
 Daniel Hiltgen
|
9b5a3c5991
Merge pull request #3914 from dhiltgen/mac_perf
|
1 year ago |
|
Jeffrey Morgan
|
00b0699c75
Reload model if `num_gpu` changes (#3920)
|
1 year ago |
|
Daniel Hiltgen
|
b123be5b71
Adjust context size for parallelism
|
1 year ago |
 Bryce Reitano
|
36a6daccab
Restructure loading conditional chain
|
1 year ago |
|
Bryce Reitano
|
284e02bed0
Move ggml loading to when we attempt fitting
|
1 year ago |
