19bc49de5f
try without version number
2025-09-20 10:48:18 +02:00
a7557cf1a8
trying again
2025-09-20 10:39:05 +02:00
3ccc18f1e1
try again
2025-09-20 10:36:48 +02:00
79a0f526b1
fixed vulkan-sdk name
2025-09-20 10:33:23 +02:00
0f86789808
fix version
2025-09-20 10:31:44 +02:00
62a8d66002
trying again
2025-09-20 10:30:31 +02:00
26df69a025
trying again
2025-09-20 10:24:31 +02:00
475d2c2583
trying to fix
2025-09-20 10:15:29 +02:00
c91b494a8b
fix version
2025-09-20 10:10:10 +02:00
af50fd5af7
try again linux build
2025-09-20 10:08:24 +02:00
236c274017
temporarly disable cuda and rocm
2025-09-20 10:00:14 +02:00
e29bb17613
trying to build vulkan for linux
2025-09-20 09:58:31 +02:00
a0389785c7
revert windows-latest
2025-09-20 09:45:36 +02:00
b244c9f9f3
revert debugging changes (vulkan builds on windows)
2025-09-20 09:44:09 +02:00
6e310d1cb6
fixed install command
2025-09-20 09:37:25 +02:00
b4595f0022
correct vulkan silent install
2025-09-20 09:31:58 +02:00
7e161f1dbf
correct vulkan install
2025-09-20 09:16:54 +02:00
d1125ea349
comment out cude for faster turnaround
2025-09-20 09:14:02 +02:00
c972cf6d46
set vulkan path
2025-09-20 09:12:14 +02:00
45f7850e75
temporarly commenting out rocm
2025-09-20 09:04:30 +02:00
e2b38c391b
commenting out error action stop
2025-09-20 09:02:55 +02:00
ed03bb7928
reenable cpu
2025-09-20 09:01:25 +02:00
c84ac53579
Commenting out other presets to build vulkan
2025-09-20 09:00:26 +02:00
a4461bc0d4
use temporarly windows-latest for build
2025-09-20 08:46:59 +02:00
6bbc054705
temporarly comment out gate to run windows task
2025-09-20 08:35:58 +02:00
0f543fdb1e
Vulkan on Windows Test
2025-09-20 08:04:11 +02:00
d5dab2d186
Merge remote-tracking branch 'upstream/main' into vulkanV3
2025-09-19 22:29:52 +02:00
62b2265f9d
buildvulkanAsSeperateFunction
2025-09-19 06:52:05 +02:00
9f3a37fd36
fix: model load for unsupported embedding models ( #12311 )
...
with #12181 , there's now support for embeddings in ollama engine.
this is done by mutating the architecture and adding _embed when it
detects an embedding model. however this introduced a bug where if
an embedding model was run based on an existing ollama engine model
without an embedding implementation, e.g. llama4, it will pass the
initial arch support check but fail when actually loaded.
there's currently two entrypoints to creating a model. previously this
second entrypoint was necessary because calling model.New would also
load the model. since #11818 , this is no longer th case so merge them
to reduce complexity
2025-09-18 16:11:08 -07:00
7460259eb3
feat: qwen3 embed ( #12301 )
...
* cleanup
* use pooling.TypeNone
* pooling test
* qwen3 embed
2025-09-18 15:50:32 -07:00
22ccdd74c2
server: add unauthorized error to remote chat handler ( #12338 )
2025-09-18 15:40:31 -07:00
0c3d0e7533
build: avoid unbounded parallel builds ( #12319 )
...
With the addition of cuda v13, on a clean setup, the level of parallelism
was causing docker desktop to become overwhelmed and compilers
were crashing. This limits to 8 parallel per build stage, with the ability
to override if you have many more cores available.
2025-09-18 14:57:01 -07:00
eb0a5d4459
auth: check the permissions on the private key to see if it's readable ( #12336 )
2025-09-18 14:34:34 -07:00
ceac416ec2
fix(integration): check truncated length ( #12337 )
2025-09-18 14:00:21 -07:00
01d8466dd6
Merge branch 'vulkanV3' of https://github.com/inforithmics/ollama into vulkanV3
2025-09-18 07:53:50 +02:00
59a83bd97e
Merge remote-tracking branch 'upstream/main' into vulkanV3
2025-09-18 07:47:40 +02:00
84257d8a7f
Merge pull request #6 from rillomas/fix-vulkan-header
...
Fixed Vulkan header
2025-09-18 07:47:07 +02:00
2717dce6fe
convert: convert bf16 vision weights to fp16 ( #12324 )
...
This change moves back to converting bf16 vision weights to fp16,
specifically if they start with the name "v." (such as v.blk.0.attn_k.weight).
This fixes a bug where converted images are failing because they are trying
to call `im2col` which doesn't have a bf16 kernel in ggml.
2025-09-17 17:43:17 -07:00
d0b5247084
Fixed Vulkan header
...
More aligned with official header definition now
2025-09-18 08:40:52 +09:00
9b8187b487
server: skip parsing initial <think> if provided in the prompt for /api/generate ( #12289 )
2025-09-17 16:39:04 -07:00
8b894933a7
engine: add remote proxy ( #12307 )
2025-09-17 14:40:53 -07:00
15eef5cc87
Merge remote-tracking branch 'upstream/main' into vulkanV3
2025-09-17 23:06:02 +02:00
5ed727815e
Merge pull request #5 from rillomas/remove-vulkan-header
...
Fix CI related errors
2025-09-17 23:05:04 +02:00
9c5bf342bc
fix: multi-cuda version skew ( #12318 )
...
Ensure that in a version skewed multi-cuda setup we use the lowest version for all GPUs
2025-09-17 13:05:09 -07:00
564b558c92
fix(llama): other llama flavours ( #12308 )
...
* fix(llama): rope scale
* spm llama
* skip moe models
* cleanup
2025-09-17 12:12:21 -07:00
a417ac97ee
prefer ollama engine for qwen3 ( #12310 )
2025-09-17 09:48:21 -07:00
ac9d59cf69
Fixed wrong structure ID
2025-09-17 16:59:23 +09:00
45430ded4b
Fixed missing members in Vulkan header
...
also added zero clear for some structs
2025-09-17 16:04:43 +09:00
6cf4e0a7c8
added missing NL
2025-09-17 15:21:24 +09:00
73441c9780
Removed unneeded function call
...
Somehow removing this call fixed the crashing when Vulkan header was removed
2025-09-17 15:11:13 +09:00
Inforithmics
Michael Yang
Jeffrey Morgan
Daniel Hiltgen
Patrick Devine
Nakasaka, Masato
frob