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

use same struct size for cuda and non cuda (+1 squashed commits)

Browse source 
Squashed commits:

[6eee8e2f] use same struct size for cuda and non cuda
This commit is contained in:
Concedo 2024-01-03 16:03:36 +08:00
parent 94e68fe474
commit e49d398f73
2 changed files with 4 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

2
gpttype_adapter.cpp
View file

@ -1532,7 +1532,7 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
} }
if (kcpp_params->seed <= 0 || kcpp_params->seed==0xFFFFFFFF) if (kcpp_params->seed <= 0 || kcpp_params->seed==0xFFFFFFFF)
{ {
kcpp_params->seed = (((uint32_t)time(NULL)) % 1000000); kcpp_params->seed = (((uint32_t)time(NULL)) % 1000000u);
} }
// tokenize the prompt // tokenize the prompt

5
llama.h
View file

@ -4,9 +4,10 @@
#include "ggml.h" #include "ggml.h"
#ifdef GGML_USE_CUBLAS #ifdef GGML_USE_CUBLAS
#include "ggml-cuda.h" #include "ggml-cuda.h"
#define LLAMA_MAX_DEVICES GGML_CUDA_MAX_DEVICES #define LLAMA_MAX_DEVICES 16
#else #else
#define LLAMA_MAX_DEVICES 1 //just max it out, same as GGML_CUDA_MAX_DEVICES
#define LLAMA_MAX_DEVICES 16
#endif // GGML_USE_CUBLAS #endif // GGML_USE_CUBLAS
#include <stddef.h> #include <stddef.h>
#include <stdint.h> #include <stdint.h>