[Benchmark] Fix custom prompt issue.
Fixed the params of TopKMethod in MoE (
#131 )
Bump torch from 2.3.0+cpu.cxx11.abi to 2.6.0
[Benchmark] Format output.
Duyi-Wangpushed 2 commits to main • ac33db4…e23103a • 11 days ago
[Qwen3] Add Qwen3 model support. (
#122 )
Duyi-Wangpushed 3 commits to main • 477f5ce…ac33db4 • 18 days ago
[Docs] Update deepseek usage. (
#121 )
Duyi-Wangpushed 2 commits to main • 7669be9…477f5ce • 19 days ago
[Example] Update qwen2 7b/14b config from qwen1.5 to 2.5. (
#125 )
[Util] Rename ShmReduction to ShmCCL (
#123 )
[Kernel] Update xDNN (optimize perf on GNR) (
#120 )
[Kernel] update xDNN (expose xDNN env) (
#119 )
[Demo] apply_chat_template in demo.py when chat=true. (
#113 )
[Benchmark] Change numactl -p to -m. (
#117 )
[Benchmark] Remove deepseek-coder-33b config which is conflict with d…
[DeepSeek] Default loading moe gate bias when noaux_tc. (
#111 )
[README] Update README. (
#107 )
[Docs] Add DeepSeek usage. (
#104 )
[Benchmark] Fix cross numa bind for GNR 128cores. (
#102 )
[Convert] Change Deepseek default dtype from bf16 to fp8.
[Dependency] Update transformers to 4.48.3. (
#100 )
[Benchmark] Add 2-Socket SNC3 mode. (
#55 )
Add condition to switch MOE Engine (
#90 )
[Demo] Add thinking process for demo (
#492 )
Pull request merge
Bump transformers from 4.40.0 to 4.48.0
[Model] Add mixtral model support (
#50 )
[Benchmark] Fix bugs in mpirun commands (
#487 )
Pull request merge
[Benchmark] Fix bug for EMR SNC-2 mode benchmark (
#484 )
Pull request merge
You can’t perform that action at this time.