thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

llama: use FA + max. GPU layers by default (#15434)

Browse source 
* llama: use max. GPU layers by default, auto -fa

* ggml-backend: abort instead of segfault
This commit is contained in:
Johannes Gäßler 2025-08-30 16:32:10 +02:00 • committed by GitHub
parent 38ad381f9f
commit e81b8e4b7f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
19 changed files with 235 additions and 72 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/llama-impl.h
View file

@ -59,3 +59,5 @@ std::string llama_format_tensor_shape(const std::vector<int64_t> & ne);
std::string llama_format_tensor_shape(const struct ggml_tensor * t);
std::string gguf_kv_to_str(const struct gguf_context * ctx_gguf, int i);
#define LLAMA_TENSOR_NAME_FATTN "__fattn__"