Code

Add qwen2moe#6074

Code

Merged

ggerganov merged 11 commits intoggml-org:masterggml-org/llama.cpp:masterfrom

simonJJJ:add_qwen2moesimonJJJ/llama.cpp:add_qwen2moeCopy head branch name to clipboard

Apr 16, 2024

Commits

Commits on Mar 15, 2024

• support qwen2moe

  simonJJJ

  committed
  56a38c4

  Copy full SHA for 56a38c4

• resolve-conflicts

  simonJJJ

  committed
  43c2c13

  Copy full SHA for 43c2c13

Commits on Apr 15, 2024

• merge qwen2moe

  simonJJJ

  committed
  edd8e2e

  Copy full SHA for edd8e2e

• fix-review

  simonJJJ

  committed
  4256fe6

  Copy full SHA for 4256fe6

• metal : support unary ops for nelements % 4 != 0

  ggerganov

  committed
  70482f5

  Copy full SHA for 70482f5

• metal : require contiguousness for float4 unary kernels

  ggerganov

  committed
  7c1ab98

  Copy full SHA for 7c1ab98

• metal : require contiguousness for float4 unary kernels (cont)

  ggerganov

  committed
  0010280

  Copy full SHA for 0010280

Commits on Apr 16, 2024

• fix-review

  simonJJJ

  committed
  7355ca8

  Copy full SHA for 7355ca8

• names : for brevity "SHARED_EXP" -> "SHEXP"

  ggerganov

  committed
  f88e684

  Copy full SHA for f88e684

• llama : reuse build_moe_ffn()

  ggerganov

  committed
  de35551

  Copy full SHA for de35551

• llama : add model type name

  ggerganov

  committed
  245565f

  Copy full SHA for 245565f