Daniel Hiltgen 359b15a597 Handle models with divergent layer sizes ...

The recent refactoring of the memory prediction assumed all layers
are the same size, but for some models (like deepseek-coder-v2) this
is not the case, so our predictions were significantly off.

2024-06-18 11:05:34 -07:00

..

ext_server

Fix server.cpp for the new cuda build macros

2024-06-14 14:51:40 -07:00

generate

Add back lower level parallel flags

2024-06-17 13:44:46 -07:00

llama.cpp @ 7c26775adb

llm: update llama.cpp commit to 7c26775 (#4896)

2024-06-17 15:56:16 -04:00

patches

llm: update llama.cpp commit to 7c26775 (#4896)

2024-06-17 15:56:16 -04:00

filetype.go

Add support for IQ1_S, IQ3_S, IQ2_S, IQ4_XS. IQ4_NL (#4322)

2024-05-23 13:21:49 -07:00

ggla.go

simplify safetensors reading

2024-05-21 11:28:22 -07:00

ggml.go

Improve multi-gpu handling at the limit

2024-06-14 14:51:40 -07:00

gguf.go

Revert "Merge pull request #4938 from ollama/mxyng/fix-byte-order"

2024-06-11 15:56:17 -07:00

llm_darwin_amd64.go

…

llm_darwin_arm64.go

…

llm_linux.go

…

llm_windows.go

…

llm.go

revert tokenize ffi (#4761)

2024-05-31 18:54:21 -07:00

memory_test.go

review comments and coverage

2024-06-14 14:55:50 -07:00

memory.go

Handle models with divergent layer sizes

2024-06-18 11:05:34 -07:00

payload.go

review comments and coverage

2024-06-14 14:55:50 -07:00

server.go

Tighten up memory prediction logging

2024-06-18 09:15:35 -07:00

status.go

…