karpathy/llm.c

Issues

• Token out of vocabulary at train_gpt2.cu:675

  #786 opened a month ago by aidando73
  1

• hellaswag.py - Github Access to this site has been restricted.

  #785 opened a month ago by aidando73
  1

• MPI run with 8 GPU fails

  #727 opened 5 months ago by msharmavikram
  2

• [cudnn_frontend] Error: No execution plans support the graph.

  #761 opened 3 months ago by Necktwi
  2

• Is there any way to make customized dataset?

  #777 opened 2 months ago by dongrixinyu
  0

• Online Softmax is wrong

  #776 opened 3 months ago by NoSavedDATA
  0

• Makefile incorrectly finds that `nccl` is installed for Linux systems with `libvncclclient`

  #774 opened 3 months ago by leiDnedyA
  0

• Problem when debugging cuda kernel functions.

  #768 opened 3 months ago by dongrixinyu
  3

• Question concerning `float4`

  #767 opened 3 months ago by dongrixinyu
  0

• Will this repo update new documentation later?

  #759 opened 3 months ago by dongrixinyu
  0

• llm.c for inference

  #752 opened 4 months ago by ztachip
  2

• Error no instance of overloaded function "..." matches the argument list

  #749 opened 4 months ago by drzsdrtfg
  1

• Suggestion: Test more Activation Functions

  #739 opened 4 months ago by linux-leo
  0

• Can't train in FP16 on Turing

  #747 opened 4 months ago by jafioti
  1

• MPI run error

  #729 opened 4 months ago by wzzanthony
  0

• TypeError: normal_() got an unexpected keyword argument 'generator'

  #723 opened 5 months ago by StarHtimE
  1

• Pretraining (with CPUs)

  #660 opened 6 months ago by bitmarkcc
  5

• How to do Inference on the trained weight of GPT 2 model after finishing the training on CPU using train_gpt2.py and train_gpt2 ?

  #372 opened 8 months ago by asifshaikat
  1

• Suggestion: Use smollm corpus

  #695 opened 5 months ago by linux-leo
  3

• Cudnn error cudnn_att.cpp on train_gptcu

  #492 opened 6 months ago by maderix
  5

• Windows issue with Cuda Toolkit 12.5 and latest MSVC compiler 17.10

  #642 opened 5 months ago by rosslwheeler
  2

• Different batch_size results in different evaluation loss.

  #710 opened 5 months ago by iminfine
  0

• Larger Tokenizers

  #701 opened 5 months ago by dustinwloring1988
  0

• Getting "Floating point exception (core dumped)" Error

  #687 opened 5 months ago by alvins82
  4

• image-gpt

  #697 opened 5 months ago by bil-ash
  1

• Is Multi-GPU config enabled even when I'm using one GPU?

  #692 opened 5 months ago by BlaiseMuhirwa
  1

• Specify torch version number in requirements.txt ?

  #656 opened 6 months ago by Phil-U-U
  2

• LLM.c in google colab

  #562 opened 6 months ago by Eliah7
  1

• Modal script - benchmarking, profiling and libraries

  #504 opened 7 months ago by vyom1611
  6

• sel4 + llm.c > path to putting these llms in any mission critical system

  #622 opened 6 months ago by torrmal
  0

• BitNet (b1.58) support

  #485 opened 7 months ago by EwoutH
  2

• Broader vendor support for hardware acceleration

  #400 opened 7 months ago by ttraenkler
  5

• Is there a plan to support 8bits (FP8 or INT8)?

  #391 opened 7 months ago by ifromeast
  2

• is max_seq_len configurable or hardcoded parameter?

  #569 opened 6 months ago by morphpiece
  2

• OSError: Memory mapping file failed: Cannot allocate memory

  #566 opened 6 months ago by antonkratz
  1

• more detailed explanation of Multi GPU

  #373 opened 7 months ago by hafezmg48
  4

• `make` fails to autodetect GPU compute capability

  #387 opened 7 months ago by aaakulchyk
  3

• Running `quick start on CPU` on Macbook Pro M2

  #563 opened 6 months ago by full-stack-ai
  7

• apparent compatibility issues with earlier c++ versions after recent pushes

  #555 opened 7 months ago by hafezmg48
  3

• I can not understand the `cublasGemmStridedBatchedEx` call in the `attention_forward`

  #557 opened 7 months ago by echosprint
  0

• ERROR on the AMD GPU

  #527 opened 7 months ago by Lookforworld
  4

• Model Export & Inference

  #502 opened 7 months ago by karpathy
  3

• Recalculating the activations in the backwards pass to conserve memory

  #478 opened 7 months ago by ChrisDryden
  3

• Mismatch of dweight at layernorm_backward.cu

  #428 opened 7 months ago by foreverpiano
  0

• Deleting Conda/Python as a dependency entirely to dramatically decrease "latency to step"

  #482 opened 7 months ago by karpathy
  4

• python dev/data/fineweb.py --version 10B

  #484 opened 7 months ago by bigsnarfdude
  2

• 2D and 3D tile divisions so that permutation coordinates can be read from threadIdx and blockIdx

  #406 opened 7 months ago by ChrisDryden
  3

• ThunderKittens Backend

  #407 opened 7 months ago by AndreSlavescu
  1

• compute sanitizers

  #393 opened 7 months ago by ngc92
  1

• Llm on small models

  #379 opened 7 months ago by Mihir0567
  0