Commits

Branch selector

 master

User selector

All users

All time

Commit History

Commits on Sep 23, 2024

• Add MI300A and adjust MI300X cores

  anthonix

  committed
  f3d8550

  Copy full SHA for f3d8550

Commits on Sep 21, 2024

• Merge remote-tracking branch 'upstream/master'

  anthonix

  committed
  a8771b5

  Copy full SHA for a8771b5

• Add gfx942 to Makefile

  anthonix

  committed
  f0bcc4c

  Copy full SHA for f0bcc4c

• Added MI300x to mfu calcs

  anthonix

  committed
  9d7b5ad

  Copy full SHA for 9d7b5ad

• Limit global norm grid size to max of 512

  anthonix

  committed
  6826f64

  Copy full SHA for 6826f64

Commits on Sep 13, 2024

• change default params: use tinyshakespeare and decrease LR

  karpathy

  committed
  bd8c604

  Copy full SHA for bd8c604

Commits on Aug 26, 2024

• Merge pull request #724 from GaoYusong/master

  Show description for bd457aa
  karpathy

  authored
  bd457aa

  Copy full SHA for bd457aa

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

  Show description for a2bdae2
  karpathy

  authored
  a2bdae2

  Copy full SHA for a2bdae2

• Merge pull request #735 from gordicaleksa/minor_refactor3

  Show description for ebc28b9
  karpathy

  authored
  ebc28b9

  Copy full SHA for ebc28b9

Commits on Aug 19, 2024

• Merge remote-tracking branch 'upstream/master'

  anthonix

  committed
  fcdcf27

  Copy full SHA for fcdcf27

Commits on Aug 16, 2024

• Merge pull request #744 from dengl11/pr

  Show description for 2c92137
  karpathy

  authored
  2c92137

  Copy full SHA for 2c92137

• Merge pull request #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

• Update README.md

  anthonix

  committed
  8fc17e6

  Copy full SHA for 8fc17e6

• Merge remote-tracking branch 'upstream/master'

  anthonix

  committed
  3b91d75

  Copy full SHA for 3b91d75

• fix a typo

  dengl11

  committed
  ff356e4

  Copy full SHA for ff356e4

Commits on Aug 13, 2024

• Merge pull request #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 #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

Pagination

Previous

Next