mirror of
https://github.com/ollama/ollama.git
synced 2025-03-29 11:11:47 +01:00
Wire up system info log for new engine (#9123)
This commit is contained in:
parent
010313bb63
commit
df2680b4b9
@ -305,6 +305,10 @@ func (b *testBackend) NewContext() ml.Context {
|
||||
return &testContext{}
|
||||
}
|
||||
|
||||
func (b *testBackend) SystemInfo() string {
|
||||
return "not implemented"
|
||||
}
|
||||
|
||||
type testContext struct{}
|
||||
|
||||
func (c *testContext) Zeros(dtype ml.DType, shape ...int) ml.Tensor {
|
||||
|
@ -23,6 +23,7 @@ type Backend interface {
|
||||
Config() Config
|
||||
Get(name string) Tensor
|
||||
NewContext() Context
|
||||
SystemInfo() string
|
||||
}
|
||||
|
||||
var backends = make(map[string]func(*os.File) (Backend, error))
|
||||
|
@ -1,11 +1,27 @@
|
||||
package ggml
|
||||
|
||||
// #cgo CPPFLAGS: -I${SRCDIR}/ggml/include
|
||||
// #include <stdlib.h>
|
||||
// #include <stdint.h>
|
||||
// #include "ggml.h"
|
||||
// #include "ggml-cpu.h"
|
||||
// #include "ggml-backend.h"
|
||||
/*
|
||||
#cgo CPPFLAGS: -I${SRCDIR}/ggml/include
|
||||
#include <stdlib.h>
|
||||
#include <stdint.h>
|
||||
#include "ggml.h"
|
||||
#include "ggml-cpu.h"
|
||||
#include "ggml-backend.h"
|
||||
static struct ggml_backend_feature * getBackendFeatures(void *fp, ggml_backend_reg_t reg) {return ((ggml_backend_get_features_t)(fp))(reg);}
|
||||
static struct ggml_backend_feature * getNextBackendFeatures(struct ggml_backend_feature * feature) { return &feature[1];}
|
||||
|
||||
typedef enum {COMP_UNKNOWN,COMP_GCC,COMP_CLANG} COMPILER;
|
||||
COMPILER inline get_compiler() {
|
||||
#if defined(__clang__)
|
||||
return COMP_CLANG;
|
||||
#elif defined(__GNUC__)
|
||||
return COMP_GCC;
|
||||
#else
|
||||
return UNKNOWN_COMPILER;
|
||||
#endif
|
||||
}
|
||||
|
||||
*/
|
||||
import "C"
|
||||
|
||||
import (
|
||||
@ -626,3 +642,34 @@ func (t *Tensor) Conv2D(ctx ml.Context, t2 ml.Tensor, s0, s1, p0, p1, d0, d1 int
|
||||
t: C.ggml_conv_2d(ctx.(*Context).ctx, t.t, t2.(*Tensor).t, C.int(s0), C.int(s1), C.int(p0), C.int(p1), C.int(d0), C.int(d1)),
|
||||
}
|
||||
}
|
||||
|
||||
func (b *Backend) SystemInfo() string {
|
||||
var compiler string
|
||||
switch C.get_compiler() {
|
||||
case C.COMP_UNKNOWN:
|
||||
compiler = "cgo(unknown_compiler)"
|
||||
case C.COMP_GCC:
|
||||
compiler = "cgo(gcc)"
|
||||
case C.COMP_CLANG:
|
||||
compiler = "cgo(clang)"
|
||||
}
|
||||
|
||||
var s string
|
||||
for i := range C.ggml_backend_reg_count() {
|
||||
reg := C.ggml_backend_reg_get(i)
|
||||
fName := C.CString("ggml_backend_get_features")
|
||||
defer C.free(unsafe.Pointer(fName))
|
||||
get_features_fn := C.ggml_backend_reg_get_proc_address(reg, fName)
|
||||
if get_features_fn != nil {
|
||||
s += C.GoString(C.ggml_backend_reg_name(reg))
|
||||
s += " : "
|
||||
for features := C.getBackendFeatures(get_features_fn, reg); features.name != nil; features = C.getNextBackendFeatures(features) {
|
||||
s += C.GoString(features.name)
|
||||
s += " = "
|
||||
s += C.GoString(features.value)
|
||||
s += " | "
|
||||
}
|
||||
}
|
||||
}
|
||||
return s + compiler
|
||||
}
|
||||
|
@ -813,6 +813,8 @@ func (s *Server) loadModel(
|
||||
panic(err)
|
||||
}
|
||||
|
||||
slog.Info("system", "info", s.model.Backend().SystemInfo() /* "threads", *threads */)
|
||||
|
||||
// TODO(jessegross): LoRA loading
|
||||
if lpath.String() != "" {
|
||||
panic("loras are not yet implemented")
|
||||
@ -881,7 +883,6 @@ func Execute(args []string) error {
|
||||
})
|
||||
slog.SetDefault(slog.New(handler))
|
||||
slog.Info("starting ollama engine")
|
||||
// TODO(jessegross): Some system info would be useful
|
||||
|
||||
server := &Server{
|
||||
batchSize: *batchSize,
|
||||
|
Loading…
x
Reference in New Issue
Block a user