Skip to content

Activity

Showing most recent first

Fix: Prioritize default head_dim when provided by architecture (Gemma…

turboderppushed 1 commit to dev • 385a516…a03db45 • 
7 days ago

Fix: Correctly read query_pre_attn_scalar from text_config (Gemma3)

turboderppushed 1 commit to dev • 17762c1…385a516 • 
7 days ago

Merge remote-tracking branch 'origin/dev' into dev

turboderppushed 17 commits to dev • eaf8ad1…17762c1 • 
8 days ago

Update chat.py, include multi-line input support and context clearing…

Pull request merge
turboderppushed 1 commit to dev • d8fa1a8…eaf8ad1 • 
12 days ago

Support partial_rotary_factor (Phi-4 mini)

turboderppushed 1 commit to dev • 2e630ae…d8fa1a8 • 
23 days ago

Fix alt pos embeddings and block diagonal mask when flash-attn is dis…

turboderppushed 1 commit to dev • 6e4a84a…2e630ae • 
on Feb 13

Update build actions

turboderppushed 1 commit to master • f1c4126…1a80d38 • 
on Feb 8

Update build actions

turboderppushed 1 commit to master • f98a7b7…f1c4126 • 
on Feb 8

Update build actions

turboderppushed 1 commit to master • 096076b…f98a7b7 • 
on Feb 8

Update build actions

turboderppushed 1 commit to master • 0f4a9f0…096076b • 
on Feb 8

Update build actions

turboderppushed 1 commit to master • f3de3cb…0f4a9f0 • 
on Feb 8

Update build actions

turboderppushed 1 commit to master • 94e5790…f3de3cb • 
on Feb 8

Update build actions

turboderppushed 1 commit to master • 3a9618d…94e5790 • 
on Feb 7

Update build actions

turboderppushed 11 commits to master • b9c025b…3a9618d • 
on Feb 7

Bump to 0.2.8

turboderppushed 1 commit to dev • d05fbcc…6e4a84a • 
on Feb 7

Fix Pixtral regression

turboderppushed 1 commit to dev • 96b2f9d…d05fbcc • 
on Feb 4

Add Qwen2.5 mode to grounding demo

turboderppushed 1 commit to dev • cce6f95…96b2f9d • 
on Jan 29

Initial support for Qwen2.5-VL

turboderppushed 1 commit to dev • d0413b0…cce6f95 • 
on Jan 29

Check length of gpu_split in model_init

turboderppushed 1 commit to dev • c8fa853…d0413b0 • 
on Jan 9

Test script: Allow --eval_rows in wiki2 ppl test

turboderppushed 4 commits to dev • ae241a9…c8fa853 • 
on Jan 9

Enable large runner

turboderppushed 1 commit to master • c41acd5…b9c025b • 
on Dec 30, 2024

Extra ROCm 6.2 actions

turboderppushed 1 commit to master • 7c08c6d…c41acd5 • 
on Dec 30, 2024

Deactivate mamba

turboderppushed 1 commit to master • c8075ca…7c08c6d • 
on Dec 30, 2024

Update conda-incubator

turboderppushed 1 commit to master • ae241a9…c8075ca • 
on Dec 30, 2024

Fix video example

turboderppushed 20 commits to master • 4f83f52…ae241a9 • 
on Dec 30, 2024

Fix video example

turboderppushed 1 commit to dev • 1ef6183…ae241a9 • 
on Dec 30, 2024

Bump to v0.2.7

turboderppushed 1 commit to dev • b010cb9…1ef6183 • 
on Dec 30, 2024

Fix compilation errors on aarch64

turboderppushed 1 commit to dev • fb5000a…b010cb9 • 
on Dec 29, 2024

Don't compile AVX2 functions when building without AVX2 support

turboderppushed 1 commit to dev • 82bb648…fb5000a • 
on Dec 29, 2024

Fix Granite3 logit scaling

turboderppushed 1 commit to dev • bee449d…82bb648 • 
on Dec 27, 2024