karpathy / llm.c

LLM training in simple, raw C/CUDA

karpathy/llm.c Issues

Suggestion: Use smollm corpus
Updated 2 days ago2
Larger Tokenizers
Updated 4 days ago
Getting "Floating point exception (core dumped)" Error
Updated 5 days ago4
image-gpt
Updated 5 days ago1
Is Multi-GPU config enabled even when I'm using one GPU?
Closed 5 days ago1
Windows issue with Cuda Toolkit 12.5 and latest MSVC compiler 17.10
Updated 6 days ago1
Pretraining (with CPUs)
Updated 15 days ago4
Specify torch version number in requirements.txt ?
Updated 18 days ago2
LLM.c in google colab
Closed 19 days ago1
Modal script - benchmarking, profiling and libraries
Updated 23 days ago6
sel4 + llm.c > path to putting these llms in any mission critical system
Closed 24 days ago
BitNet (b1.58) support
Updated 25 days ago2
Broader vendor support for hardware acceleration
Updated a month ago5
Is there a plan to support 8bits (FP8 or INT8)?
Updated a month ago2
is max_seq_len configurable or hardcoded parameter?
Updated a month ago2
Error: make: *** [Makefile:203: train_gpt2cu] Error 255
Updated a month ago6
OSError: Memory mapping file failed: Cannot allocate memory
Closed 2 months ago1
Cudnn error cudnn_att.cpp on train_gptcu
Closed 2 months ago4
void tokenizer_init failed
Closed 2 months ago3
Possible bugs in the data loading functions
Updated 2 months ago1
more detailed explanation of Multi GPU
Updated 2 months ago4
`make` fails to autodetect GPU compute capability
Updated 2 months ago3
Running `quick start on CPU` on Macbook Pro M2
Closed 2 months ago7
inf loss at big batch
Updated 2 months ago1
Assertion `graph->check_support(cudnn_handle).is_good()' failed
Updated 2 months ago20
apparent compatibility issues with earlier c++ versions after recent pushes
Closed 2 months ago3
I can not understand the `cublasGemmStridedBatchedEx` call in the `attention_forward`
Closed 2 months ago
ERROR on the AMD GPU
Closed 2 months ago4
Model Export & Inference
Updated 2 months ago3
Recalculating the activations in the backwards pass to conserve memory
Updated 2 months ago3
Mismatch of dweight at layernorm_backward.cu
Closed 2 months ago
Deleting Conda/Python as a dependency entirely to dramatically decrease "latency to step"
Updated 2 months ago4
python dev/data/fineweb.py --version 10B
Closed 2 months ago2
2D and 3D tile divisions so that permutation coordinates can be read from threadIdx and blockIdx
Updated 2 months ago3
ThunderKittens Backend
Updated 2 months ago1
WikiText 103 evaluation
Closed 2 months ago1
compute sanitizers
Updated 2 months ago1
MultiGPU training hangs
Closed 2 months ago9
Llm on small models
Closed 2 months ago
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 ?
Updated 3 months ago
Hardcoded block_size in kernels
Updated 3 months ago14
make: *** [Makefile:194: train_gpt2] Error 2 on Windows
Closed 3 months ago6
What would be the main design trade-offs when re-implementing in clean modern C++?
Updated 3 months ago2
About pull request of custom kernel implementation
Closed 3 months ago
When will llama and other frameworks be supported?
Updated 3 months ago1
delete use of cooperative groups in kernels
Updated 3 months ago2
Possible NULL Pointer Dereference
Closed 3 months ago2
cuda code that approaches cublas performance
Updated 3 months ago
Refactoring all of the shared cuda helper methods to the shared common file
Closed 3 months ago
Splitting cuda dev files to use smaller sizes for cpu validation compared to profiling
Updated 3 months ago