Commits

Breadcrumbs

History for

on3cc16f1

User selector

All users

Datepicker

All time

Commit History

Commits on May 5, 2024

• Minor update on the code

  lancerts

  committed
  bbfe8c9

Commits on May 3, 2024

• move gelu_backward to backward block

  lancerts

  committed
  c81adeb

Commits on May 2, 2024

• small doc fixes

  karpathy

  committed
  937fbd8

• bias backward kernel that will use all available threads

  ngc92

  committed
  6e48501

• pragma unroll quick fix

  karpathy

  committed
  50714d2

• update encoder_forward with the latest and greatest Packed128 and incorporate into train_gpt2cu

  karpathy

  committed
  9da2729

• Merge branch 'patch-1' of https://github.com/JaneIllario/llm.c into JaneIllario-patch-1

  karpathy

  committed
  3d1761b

• Merge branch 'gelu_backwards' of https://github.com/JaneIllario/llm.c into JaneIllario-gelu_backwards-3

  karpathy

  committed
  99f51ba

• Merge branch 'removing_cg_example' of https://github.com/ChrisDryden/llm.c into ChrisDryden-removing_cg_example

  karpathy

  committed
  ec73086

• add kernel 4 to docs. have to improve these docs more and document them better

  karpathy

  committed
  2202c9a

• Removed cooperative groups and added packed128 to fused classifier and prepare softmax blockwide

  

  “Christopher

  committed
  398cdaf

• Added example of removing cooperative groups

  

  “Christopher

  committed
  665d0a4

• fix typo

  JaneIllario

  committed
  d6d0d50

• update residual_forward to use util functions

  JaneIllario

  committed
  d0e5fea

• update residual forward to allow kernels to use both types

  JaneIllario

  committed
  52b4dd0

• Update residual_forward.cu

  JaneIllario

  committed
  4ffcf5b

Commits on May 1, 2024

• update kernel with util functions

  JaneIllario

  committed
  3de3c53

• update gelu backward rto allow all kernels to use both types

  JaneIllario

  committed
  7746217

• Create gelu_backward.cu

  JaneIllario

  committed
  8007280

• this kernel will be released later, sorry for the spoiler :)

  ngc92

  committed
  b42db70

• fixup

  ngc92

  committed
  91bc72d

• mixed precision utilities for dev/cuda

  ngc92

  committed
  06980b6

• Add missing cudnnDestroy() calls

  ademeure

  committed
  1147983

• attention_forward.cu ~matches train_gpt2cu & allocate less memory for stats in "att"

  ademeure

  committed
  2b540b6

• Rename lowp_float to floatX in /dev/cuda/attention_forward.cu

  ademeure

  committed
  be54842

• Merge remote-tracking branch 'karpathy/master' into cudnn_try2

  ademeure

  committed
  c4ecc04

Commits on Apr 30, 2024

• small tweaks

  karpathy

  committed
  769c911

• fix up bandwidth calculation

  ngc92

  committed
  603d862

• fixed potential error and generalized gelu forward

  ngc92

  committed
  c3515cf

• delete packed128 .fp32, use (float) ok

  karpathy

  committed
  7ba1ed8

• fixes to the PR, careful with float/floatX etc

  karpathy

  committed
  9141e04

• Merge branch 'gelu_forwards_packing' of https://github.com/ChrisDryden/llm.c into ChrisDryden-gelu_forwards_packing

  karpathy

  committed
  242981e

• add modal script

  leloykun

  committed
  b239b67

• re-commit of my old cudnn forward attention changes

  ademeure

  committed
  1c516c7

• clear the L2 cache between consecutive invokations of our microbenchmarks to get reliable results

  ngc92

  committed
  d7813d2

Pagination

Previous

Next