ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Add ReLU and SQR CUDA ops to (partially) fix Persimmon offloading (#4041)

Browse source 
* Add ReLU and SQR CUDA ops to fix Persimmon offloading

* Persimmon loader: More helpful error on CUDA/ROCM when offloading too many layers
This commit is contained in:
Kerfuffle 2023-11-13 01:58:15 -07:00 • committed by GitHub
parent 21fd874c8d
commit bb50a792ec
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 79 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

7
llama.cpp
View file

@ -2877,6 +2877,13 @@ static void llm_load_tensors(
ggml_backend_type backend_output;
if (n_gpu_layers > int(n_layer)) {
#ifdef GGML_USE_CUBLAS
if (n_gpu_layers > int(n_layer + 1)) {
LLAMA_LOG_ERROR("%s: CUDA backend missing Persimmon CUDA ops, can offload at most %ld layers. See: https://github.com/ggerganov/llama.cpp/issues/4038\n",
__func__, n_layer + 1);
throw std::runtime_error("Persimmon CUDA offload failed");
}
#endif
// norm is not performance relevant on its own but keeping it in VRAM reduces data copying
// on Windows however this is detrimental unless everything is on the GPU
#ifndef _WIN32