Code

metal gpu matrix3D addition test#87

Code

Open

DerrickYLJ wants to merge 39 commits intomainmit-han-lab/TinyChatEngine:mainfrom MetalGPUmit-han-lab/TinyChatEngine:MetalGPUCopy head branch name to clipboard

+34,379-639

Commits

Commits on Jan 15, 2024

• metal gpu matrix3D addition test

  DerrickYLJ

  committed
  4b38bb7

• metal gpu matrix3D addition test

  Show description for cb5419a
  DerrickYLJ

  committed
  cb5419a

Commits on Jan 16, 2024

• add element-wise add test in swift

  DerrickYLJ

  committed
  d23203c

Commits on Jan 24, 2024

• metal c++ version inside cpp folder

  DerrickYLJ

  committed
  3005bb3

Commits on Jan 25, 2024

• minor fix

  RaymondWang0

  committed
  7cad520

• speedup issue with metal

  DerrickYLJ

  committed
  69c929d

Commits on Jan 26, 2024

• minor fix

  RaymondWang0

  committed
  11fa7b6

Commits on Jan 27, 2024

• reorganized clear version for metal main

  DerrickYLJ

  committed
  37ff487

• matmul metal

  DerrickYLJ

  committed
  c3e3316

• matmul correctness pass

  DerrickYLJ

  committed
  d55be07

• matmul work

  DerrickYLJ

  committed
  ab446e7

Commits on Jan 28, 2024

• header param

  DerrickYLJ

  committed
  69407d5

Commits on Jan 29, 2024

• metal matmul Int4 working

  DerrickYLJ

  committed
  dd9fda7

• kernel minor change

  DerrickYLJ

  committed
  e8331c3

Commits on Jan 30, 2024

• add metal op

  RaymondWang0

  committed
  e79e702

Commits on Jan 31, 2024

• interleave two versions equal

  DerrickYLJ

  committed
  bc5cad2

• add matmulInt4_SIMD with different versions

  Show description for 3a3e97d
  DerrickYLJ

  committed
  3a3e97d

Commits on Feb 1, 2024

• building metal interface, done with kernel ops

  DerrickYLJ

  committed
  17e90b6

• update metal-cpp

  RaymondWang0

  committed
  a3f8852

• everything except for TODO

  DerrickYLJ

  committed
  246744b

• rms_norm done

  DerrickYLJ

  committed
  1b5d027

Commits on Feb 2, 2024

• done with softmax, rope and minor fix on opparam struct

  DerrickYLJ

  committed
  c5fb54f

• update for test

  DerrickYLJ

  committed
  189ed98

Commits on Feb 4, 2024

• updated matmulf32 + test script

  DerrickYLJ

  committed
  3217635

• new test script on normalization

  DerrickYLJ

  committed
  06c472a

Commits on Feb 20, 2024

• more metal ops (TODO needed)

  DerrickYLJ

  committed
  0c38582

Commits on Feb 26, 2024

• All basic operations and matmul are included

  DerrickYLJ

  committed
  962266e

Commits on Mar 11, 2024

• update llama matmul parameter and test

  DerrickYLJ

  committed
  cd416a3

Commits on Mar 22, 2024

• new format for metal in general

  DerrickYLJ

  committed
  67ea0cc

• fix matmul

  RaymondWang0

  committed
  2b71f9f

Commits on Mar 26, 2024

• add new kernels and reorganize

  DerrickYLJ

  committed
  f792a1b

• Merge branch 'MetalGPU' of https://github.com/mit-han-lab/TinyChatEngine into MetalGPU

  DerrickYLJ

  committed
  2440e9e

Commits on Apr 6, 2024

• fix parameters

  RaymondWang0

  committed
  0b094ad

Commits on Apr 16, 2024

• draft llama.cpp model

  DerrickYLJ

  committed
  926dd0d

Commits on Apr 26, 2024

• update rope

  DerrickYLJ

  committed
  3ac955c

Commits on May 1, 2024

• update makefile

  RaymondWang0

  committed
  46d9627

Commits on May 12, 2024

• reorganize metal

  DerrickYLJ

  committed
  0d4cd66

Commits on May 16, 2024

• minor fix

  DerrickYLJ

  committed
  562d32a

Commits on May 24, 2024

• add op source

  DerrickYLJ

  committed
  a269001