Code

History

Commits on move/encoder_kernel to dev/cuda35+
  1. Merge pull request #512 from gordicaleksa/refactor_encoder_bwd_kernelAndrej committed 2 years ago80e0c3b
  2. Remove redundant CPU computationAleksa Gordic committed 2 years ago290c00a
  3. Merge pull request #495 from ChrisDryden/shared_memoryAndrej committed 2 years agoa26041a
  4. Update compile cmd in the dev/cuda READMEAleksa Gordic committed 2 years ago8eafd40
  5. Removed unnecesary shared memory due to blockreduce using static defined shared memoryChristopher committed 2 years ago5450632
  6. Merge pull request #493 from vyom1611/master-2Andrej committed 2 years ago55f3665
  7. Modal benchmarking script updated to replace deprecated callsvyom1611 committed 2 years ago062f096
  8. amendlancerts committed 2 years ago47c670b
  9. spell explicitly uint to unsigned intlancerts committed 2 years agoc4c985c
  10. fix the issue Mismatch of dweight at layernorm_backward.culancerts committed 2 years agof27ca4d
  11. bugfixErik Schultheis committed 2 years ago7dc3b7d
  12. fix C % 256 != 0Erik Schultheis committed 2 years ago5f73ecf
  13. fail fast and hard; don't go into the deadlockErik Schultheis committed 2 years agod35daf1
  14. some comments and optimized shared memory amountErik Schultheis committed 2 years ago3b68082
  15. fully vectorized smem accessErik Schultheis committed 2 years ago9eba86f
  16. utilities for Packed and more vectorizationErik Schultheis committed 2 years agocb8cc25
  17. use vectorized access to shared memoryErik Schultheis committed 2 years ago53ee329
  18. Merge pull request #439 from lancerts/matmul-fixAndrej committed 2 years ago6a7fd56
  19. Merge branch 'deterministic_layernorm' of https://github.com/ademeure/llm.c into ademeure-deterministic_layernormAndrej Karpathy committed 2 years agodbacaf8
  20. int -> int64_tErik Schultheis committed 2 years ago1b98637
  21. added new layernorm backward to /dev/cuda/ademeure committed 2 years ago7cbeefc
  22. update the utils function and assertlancer committed 2 years ago2b0667a
  23. fix the unsupported block_sizelancer committed 2 years ago6348d41
  24. small touchups to grad clipAndrej Karpathy committed 2 years agoc2d12f7
  25. fixed up dev/cudaErik Schultheis committed 2 years ago66ce576
  26. added a useful mixed precision utility for dev/cudaErik Schultheis committed 2 years agod7a81ef
  27. Merge pull request #383 from KarhouTam/feature/online-softmax-forward-without-cgsAndrej committed 2 years ago109f516
  28. deterministic kernelErik Schultheis committed 2 years ago858c6e6
  29. general cleanupErik Schultheis committed 2 years ago2ccdfb7
  30. remove legacy commentAndrej Karpathy committed 2 years ago2d43e5b
  31. Merge pull request #408 from ngc92/layernorm-bw-dev-cudaAndrej committed 2 years ago222d59f
  32. update dev/cuda/layernorm_backward and improve `validate_result` to take into account fp epsilon when comparing resultsErik Schultheis committed 2 years agoe553e2f
  33. fix layernorm backward: accumulate weight gradientErik Schultheis committed 2 years agodd8c9f5
  34. fixup commentErik Schultheis committed 2 years agoc66e48c
  35. fix CI compile by disabling kernel 5Erik Schultheis committed 2 years ago65727d5