GitLab

Better example for multi-modal input

* add prompt tokens to embed response

* rm slog

* metrics

* types

* prompt n

* clean up

* reset submodule

* update tests

* test name

* list metrics

Prevent partial loading on mixed GPU brands

In mult-brand GPU setups, if we couldn't fully load the model we
would fall through the scheduler and mistakenly try to load across
a mix of brands.  This makes sure we find the set of GPU(s) that
best fit for the partial load.

* Update example models

* Remove unused README.md

Better explain multi-gpu behavior

Ensure amd gpu nodes are numerically sorted

Report better error on cuda unsupported os/arch

* hot fix

* backend stream support

* clean up

* finish reason

* move to openai

Explain font problems on windows 10

HuggingFace -> Hugging Face

fix: model save

Enable windows error dialog for subprocess

fix nil deref in auth.go

This fixes various data races scattered throughout the download/pull
client where the client was accessing the download state concurrently.

This commit is mostly a hot-fix and will be replaced by a new client one
day soon.

Also, remove the unnecessary opts argument from downloadChunk.

autodetect stop parameters from template

stop parameter is saved as a slice which is incompatible with modelfile
parsing

docs

Co-authored-by: Jeffrey Morgan <jmorganca@gmail.com>

This changes the registry client to reuse the original download URL
it gets on the first redirect response for all subsequent requests,
preventing thundering herd issues when hot new LLMs are released.

This reverts commit bb46bbcf.

If we detect an NVIDIA GPU, but nvidia doesn't support the os/arch,
this will report a better error for the user and point them to docs
to self-install the drivers if possible.

For systems that enumerate over 10 CPUs the default lexicographical
sort order interleaves CPUs and GPUs.

* float cmp

* increase tolerance