GitLab

Previously, some costly things were causing the loading of GGUF files
and their metadata and tensor information to be VERY slow:

  * Too many allocations when decoding strings
  * Hitting disk for each read of each key and value, resulting in a
    not-okay amount of syscalls/disk I/O.

The show API is now down to 33ms from 800ms+ for llama3 on a macbook pro
m3.

This commit also prevents collecting large arrays of values when
decoding GGUFs (if desired). When such keys are encountered, their
values are null, and are encoded as such in JSON.

Also, this fixes a broken test that was not encoding valid GGUF.

While models are loading, the VRAM metrics are dynamic, so try
to load on a GPU that doesn't have a model actively loading, or wait
to avoid races that lead to OOMs

This library will give us the most reliable free VRAM reporting on windows
to enable concurrent model scheduling.

Our default behavior today is to try to fit into a single GPU if possible.
Some users would prefer the old behavior of always spreading across
multiple GPUs even if the model can fit into one.  This exposes that
tunable behavior.

The spelling of the term "request" has been corrected, which was previously mistakenly written as "requeset" in the error log message.

The APIs we query are optimistic on free space, and windows pages
VRAM, so we don't have to wait to see reported usage recover on unload

Make sure the first GPU has the most free space

* dont clamp ctx size in `PredictServerFit`

* minimum 4 context

* remove context warning

The GPU drivers take a while to update their free memory reporting, so we need
to wait until the values converge with what we're expecting before proceeding
to start another runner in order to get an accurate picture.

This moves all the env var reading into one central module
and logs the loaded config once at startup which should
help in troubleshooting user server logs

This also bumps up the default to be 50 queued requests
instead of 10.

log when the waiting for the process to stop to help debug when other tasks execute during this wait.
expire timer clear the timer reference because it will not be reused.
close will clean up expireTimer if calling code has not already done this.

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.

Prior refactoring passes accidentally removed the logic to bypass VRAM
checks for CPU loads.  This adds that back, along with test coverage.

This also fixes loaded map access in the unit test to be behind the mutex which was
likely the cause of various flakes in the tests.

* reload model if `num_gpu` changes

* dont reload on -1

* fix tests

This change adds support for multiple concurrent requests, as well as
loading multiple models by spawning multiple runners. The default
settings are currently set at 1 concurrent request per model and only 1
loaded model at a time, but these can be adjusted by setting
OLLAMA_NUM_PARALLEL and OLLAMA_MAX_LOADED_MODELS.