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

hotfix for rwkv

Browse source
This commit is contained in:
Concedo 2023-06-13 23:38:41 +08:00
parent 9db2ec068f
commit 82cf97ce92
3 changed files with 3 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml.h
View file

@ -194,7 +194,7 @@
#define GGML_QNT_VERSION_FACTOR 1000 // do not change this #define GGML_QNT_VERSION_FACTOR 1000 // do not change this
#define GGML_MAX_DIMS 4 #define GGML_MAX_DIMS 4
#define GGML_MAX_NODES 4096 #define GGML_MAX_NODES 8192
#define GGML_MAX_PARAMS 256 #define GGML_MAX_PARAMS 256
#define GGML_MAX_CONTEXTS 64 #define GGML_MAX_CONTEXTS 64
#define GGML_MAX_OPT 4 #define GGML_MAX_OPT 4

2
gpttype_adapter.cpp
View file

@ -479,7 +479,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
} }
else else
{ {
n_batch = 10; //use sequence mode to speedup n_batch = 8; //use sequence mode to speedup
//setup buffers for rwkv state //setup buffers for rwkv state
auto padding = 512u; auto padding = 512u;

2
koboldcpp.py
View file

@ -224,7 +224,7 @@ maxctx = 2048
maxlen = 256 maxlen = 256
modelbusy = False modelbusy = False
defaultport = 5001 defaultport = 5001
KcppVersion = "1.30" KcppVersion = "1.30.1"
class ServerRequestHandler(http.server.SimpleHTTPRequestHandler): class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
sys_version = "" sys_version = ""