mirror of
https://github.com/ollama/ollama.git
synced 2025-03-18 22:01:47 +01:00
This provides integration with the new Ollama engine (5824541 next ollama runner (#7913)) and the rest of the Ollama infrastructure such as the runner and Ollama server. In addition, it also builds out the KV cache infrastructure to support requirements of how Ollama runs models such as: - Parallel processing - Memory management for defragmentation and shifting - Multi-modal modals Both old and new engines continue to be supported. By default, only the old engine is used. To enable the new engine: Start the server with the OLLAMA_NEW_ENGINE environment variable set: OLLAMA_NEW_ENGINE=1 ./ollama serve Start a model that is supported by the Ollama engine. This one is Llama 3.1 8b Q4_K_M: ./ollama run jessegross/llama3.1
137 lines
3.1 KiB
Go
137 lines
3.1 KiB
Go
package model
|
|
|
|
import (
|
|
"reflect"
|
|
"slices"
|
|
"testing"
|
|
|
|
"github.com/google/go-cmp/cmp"
|
|
"github.com/ollama/ollama/ml"
|
|
"github.com/ollama/ollama/ml/backend/ggml"
|
|
"github.com/ollama/ollama/ml/nn"
|
|
)
|
|
|
|
func TestParseTags(t *testing.T) {
|
|
cases := []struct {
|
|
value string
|
|
want Tag
|
|
}{
|
|
{
|
|
value: "output",
|
|
want: Tag{
|
|
Name: "output",
|
|
},
|
|
},
|
|
{
|
|
value: "output,alt:token_embd",
|
|
want: Tag{
|
|
Name: "output",
|
|
Alternate: []string{
|
|
"token_embd",
|
|
},
|
|
},
|
|
},
|
|
}
|
|
|
|
for _, tt := range cases {
|
|
t.Run(tt.value, func(t *testing.T) {
|
|
got := ParseTags(tt.value)
|
|
if diff := cmp.Diff(tt.want, got); diff != "" {
|
|
t.Errorf("ParseTags() returned unexpected values (-want +got):\n%s", diff)
|
|
}
|
|
})
|
|
}
|
|
}
|
|
|
|
type fakeBackend struct {
|
|
*ggml.Backend
|
|
names []string
|
|
}
|
|
|
|
type fakeTensor struct {
|
|
*ggml.Tensor
|
|
Name string
|
|
}
|
|
|
|
func (m *fakeBackend) Get(name string) ml.Tensor {
|
|
if slices.Contains(m.names, name) {
|
|
return &fakeTensor{Name: name}
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func TestPopulateFields(t *testing.T) {
|
|
type fakeLayer struct {
|
|
Query *nn.Linear `gguf:"attn_q"`
|
|
Key *nn.Linear `gguf:"attn_k"`
|
|
Value *nn.Linear `gguf:"attn_v"`
|
|
Output *nn.Linear `gguf:"attn_o"`
|
|
}
|
|
|
|
type fakeModel struct {
|
|
Input *nn.Embedding `gguf:"input"`
|
|
OutputNorm *nn.RMSNorm `gguf:"output_norm"`
|
|
Output *nn.Linear `gguf:"output"`
|
|
Layers [2]fakeLayer `gguf:"blk"`
|
|
}
|
|
|
|
var m fakeModel
|
|
v := reflect.ValueOf(&m)
|
|
v.Elem().Set(populateFields(Base{b: &fakeBackend{
|
|
names: []string{
|
|
"input.weight",
|
|
"blk.0.attn_q.weight",
|
|
"blk.0.attn_k.weight",
|
|
"blk.0.attn_v.weight",
|
|
"blk.1.attn_q.weight",
|
|
"blk.1.attn_k.weight",
|
|
"blk.1.attn_v.weight",
|
|
"output_norm.weight",
|
|
"output.weight",
|
|
},
|
|
}}, v.Elem()))
|
|
|
|
if diff := cmp.Diff(fakeModel{
|
|
Input: &nn.Embedding{Weight: &fakeTensor{Name: "input.weight"}},
|
|
OutputNorm: &nn.RMSNorm{Weight: &fakeTensor{Name: "output_norm.weight"}},
|
|
Output: &nn.Linear{Weight: &fakeTensor{Name: "output.weight"}},
|
|
Layers: [2]fakeLayer{
|
|
{
|
|
Query: &nn.Linear{Weight: &fakeTensor{Name: "blk.0.attn_q.weight"}},
|
|
Key: &nn.Linear{Weight: &fakeTensor{Name: "blk.0.attn_k.weight"}},
|
|
Value: &nn.Linear{Weight: &fakeTensor{Name: "blk.0.attn_v.weight"}},
|
|
},
|
|
{
|
|
Query: &nn.Linear{Weight: &fakeTensor{Name: "blk.1.attn_q.weight"}},
|
|
Key: &nn.Linear{Weight: &fakeTensor{Name: "blk.1.attn_k.weight"}},
|
|
Value: &nn.Linear{Weight: &fakeTensor{Name: "blk.1.attn_v.weight"}},
|
|
},
|
|
},
|
|
}, m); diff != "" {
|
|
t.Errorf("populateFields() set incorrect values (-want +got):\n%s", diff)
|
|
}
|
|
}
|
|
|
|
func TestPopulateFieldsAlternateName(t *testing.T) {
|
|
type fakeModel struct {
|
|
Input *nn.Embedding `gguf:"input"`
|
|
Output *nn.Linear `gguf:"output,alt:input"`
|
|
}
|
|
|
|
m := fakeModel{}
|
|
v := reflect.ValueOf(&m)
|
|
v.Elem().Set(populateFields(Base{b: &fakeBackend{
|
|
names: []string{
|
|
"input.weight",
|
|
},
|
|
}}, v.Elem()))
|
|
|
|
if diff := cmp.Diff(fakeModel{
|
|
Input: &nn.Embedding{Weight: &fakeTensor{Name: "input.weight"}},
|
|
Output: &nn.Linear{Weight: &fakeTensor{Name: "input.weight"}},
|
|
}, m); diff != "" {
|
|
t.Errorf("populateFields() set incorrect values (-want +got):\n%s", diff)
|
|
}
|
|
}
|