Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Aug 26, 2024

• Merge pull request karpathy#724 from GaoYusong/master

  Show description for bd457aa
  karpathy

  authored
  bd457aa

  Copy full SHA for bd457aa

• Merge pull request karpathy#733 from zhangpiu/feature/llm.cpp

  Show description for a2bdae2
  karpathy

  authored
  a2bdae2

  Copy full SHA for a2bdae2

• Merge pull request karpathy#735 from gordicaleksa/minor_refactor3

  Show description for ebc28b9
  karpathy

  authored
  ebc28b9

  Copy full SHA for ebc28b9

Commits on Aug 16, 2024

• Merge pull request karpathy#744 from dengl11/pr

  Show description for 2c92137
  karpathy

  authored
  2c92137

  Copy full SHA for 2c92137

• Merge pull request karpathy#745 from karpathy/feature/managed2

  Show description for 0ddedf9
  karpathy

  authored
  0ddedf9

  Copy full SHA for 0ddedf9

• i misspelled reduced

  karpathy

  committed
  18298f3

  Copy full SHA for 18298f3

• reduce across GPUs nicer

  karpathy

  committed
  8c586f9

  Copy full SHA for 8c586f9

• fallback to memory allocation of m,v,master_weights on host automatically in case of OOM. will run slower but won't OOM

  karpathy

  committed
  e6856bc

  Copy full SHA for e6856bc

• fix makefile for multigpu setups

  karpathy

  committed
  2882ec6

  Copy full SHA for 2882ec6

Commits on Aug 15, 2024

• hint to host

  ngc92

  committed
  f72c1f2

  Copy full SHA for f72c1f2

• just try to allocate on device; fallback if that fails

  ngc92

  committed
  c845757

  Copy full SHA for c845757

• fall back to cudaMallocManaged for optimizer states if we're out of memory

  ngc92

  committed
  0d52d2a

  Copy full SHA for 0d52d2a

Commits on Aug 14, 2024

• fix a typo

  dengl11

  committed
  ff356e4

  Copy full SHA for ff356e4

Commits on Aug 13, 2024

• Merge pull request karpathy#740 from karpathy/gordicaleksa-fix_dataloader2

  Show description for 4c84bc7
  karpathy

  authored
  4c84bc7

  Copy full SHA for 4c84bc7

• fix tokenizer omg

  karpathy

  committed
  755458d

  Copy full SHA for 755458d

• attempt to fix PR

  karpathy

  committed
  16635d4

  Copy full SHA for 16635d4

Commits on Aug 12, 2024

• Merge branch 'fix_dataloader2' of https://github.com/gordicaleksa/llm.c into gordicaleksa-fix_dataloader2

  karpathy

  committed
  9740a65

  Copy full SHA for 9740a65

• Merge pull request karpathy#738 from ademeure/faster_compile

  Show description for 1787210
  karpathy

  authored
  1787210

  Copy full SHA for 1787210

• Refactor tokenize, delegate instead

  gordicaleksa

  committed
  68624cc

  Copy full SHA for 68624cc

• Refactor magic/version header dict

  gordicaleksa

  committed
  e8739c8

  Copy full SHA for e8739c8

• Improve compile time via nvidia-smi instead of __nvcc_device_query (& add "--threads=0").

  ademeure

  committed
  fc88344

  Copy full SHA for fc88344

• Improve compile time via nvidia-smi instead of __nvcc_device_query (& add "--threads=0").

  ademeure

  committed
  9b1c14b

  Copy full SHA for 9b1c14b

Commits on Aug 10, 2024

• Replace fire with argparse

  gordicaleksa

  committed
  eb1b359

  Copy full SHA for eb1b359

• Add fire to reqs

  gordicaleksa

  committed
  311c149

  Copy full SHA for 311c149

• Add fineweb llama 3 support

  gordicaleksa

  committed
  5076563

  Copy full SHA for 5076563

• Use fire module, update readme

  gordicaleksa

  committed
  f8c1ff5

  Copy full SHA for f8c1ff5

• Add tiny shakespeare

  gordicaleksa

  committed
  15b76bf

  Copy full SHA for 15b76bf

• Change default dataset path to tiny stories

  gordicaleksa

  committed
  b018b0c

  Copy full SHA for b018b0c

• Update header description

  gordicaleksa

  committed
  71419cb

  Copy full SHA for 71419cb

• Add llama 3 tokenization and data loader

  gordicaleksa

  committed
  d3151c4

  Copy full SHA for d3151c4

• Refactor flash logic

  gordicaleksa

  committed
  b0bc864

  Copy full SHA for b0bc864

• Remove logprobs - like andrej in nano llama 3

  gordicaleksa

  committed
  a0c666b

  Copy full SHA for a0c666b

• Add msg to asserts

  gordicaleksa

  committed
  92cc4eb

  Copy full SHA for 92cc4eb

• Minor refactor

  gordicaleksa

  committed
  a127873

  Copy full SHA for a127873

• Update README.md

  zhangpiu

  authored
  25a302f

  Copy full SHA for 25a302f

Pagination

Previous

Next