GitLab

* perf: build graph for next batch in parallel to keep GPU busy

This refactors the main run loop of the ollama runner to perform the main GPU
intensive tasks (Compute+Floats) in a go routine so we can prepare the next
batch in parallel to reduce the amount of time the GPU stalls waiting for the
next batch of work.

* tests: tune integration tests for ollama engine

This tunes the integration tests to focus more on models supported
by the new engine.

* test: improve scheduler/concurrency stress tests

The scheduler test used to use approximate memory figures and would often
over or under shoot a systems capcity leading to flaky test results.
This should improve the reliability of this scenario by leveraging
ps output to determinie exactly how many models it takes to
trigger thrashing.

The concurrency test is also refined to target num_parallel + 1 and handle
timeouts better.

With these refinements, TestMultiModelConcurrency was redundant

* test: add parallel generate with history

TestGenerateWithHistory will help verify caching and context
are properly handled while making requests

* test: focus embed tests on embedding models

remove non-embedding models from the embedding tests

Both interface{} and any (which is just an alias for interface{} introduced in Go 1.18) represent the empty interface that all types satisfy.

Fragmentation of the KV cache can occur due to cache shifting or
different sequences getting processed. Decode uses a heuristic to
decide if it should defrag. However, this heuristic isn't 100%
accurate, so decoding can sometimes fail by surprise.

For these cases, if decode indicates that there is no KV cache space,
we should defrag and then try again.

On the smaller GPUs, the initial model load of llama2 took over 30s (the
default timeout for the DoGenerate helper)

adjust timing on some tests so they don't timeout on small/slow GPUs

Still not complete, needs some refinement to our prediction to understand the
discrete GPUs available space so we can see how many layers fit in each one
since we can't split one layer across multiple GPUs we can't treat free space
as one logical block

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.

Confirmed this fails on 0.1.30 with known regression
but passes on main

Cleaner shutdown logic, a bit of response hardening

If images aren't present, pull them.
Also fixes the expected responses

This uplevels the integration tests to run the server which can allow
testing an existing server, or a remote server.