set codesign timeout to longer (ollama#4605)

Use flash attention flag for now (ollama#4580)

* put flash attention behind flag for now

* add test

* remove print

* up timeout for sheduler tests

Merge pull request ollama#4543 from ollama/mxyng/simple-safetensors

simplify safetensors reading

fix the cpu estimatedTotal memory + get the expiry time for loading m…

…odels (ollama#4461)

Fix OpenAI `finish_reason` values when empty (ollama#4368)

Revert "only forward some env vars"

This reverts commit ce3b212.

Merge pull request ollama#4323 from dhiltgen/sort_by_free

Always use the sorted list of GPUs

omit empty done reason

Merge pull request ollama#4231 from ollama/mxyng/parser

types/model: fix parser for empty values

Merge pull request ollama#4188 from dhiltgen/use_our_lib

User our bundled libraries (cuda) instead of the host library