mirrors/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-05-11 10:26:53 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

next ollama runner (#7913)

Browse source 
feat: add new Ollama engine using ggml through cgo

This change introduces a new way to run pretrained models. It introduces 3 high level interfaces and a bunch of smaller helper interfaces to facilitate this.

- `model.Model` defines the interface for a model architecture. Models such as `llama` and `mllama`, which are provided as examples, can implement the model's forward propagation in the `Forward` method. This method will be called to generate completions. This interface can be found in `model/model.go`
- `ml.Backend` defines the interface for a backend tensor library, in this case `ggml`. Among other things, a Backend is responsible for loading a pretrained model into hardware (GPU, CPU, etc) and providing an interface for Models to access loaded tensors. This interface can be found in `ml/backend.go`
- `ml.Tensor` defines the interface for a tensor and tensor operations

This is the first implementation of the new engine. Follow up PRs will implement more features:

- non-greedy sampling (#8410)
- integration with Ollama and KV caching (#8301)
- more model support (#9080) with more coming soon

Co-authored-by: Bruce MacDonald <brucewmacdonald@gmail.com>
This commit is contained in:
Michael Yang 2025-02-14 00:31:21 +00:00 • committed by GitHub
parent 8cf16063a5
commit 58245413f4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
57 changed files with 475427 additions and 494 deletions
Download patch file Download diff file Expand all files Collapse all files

25
server/routes_generate_test.go
View file

@ -16,6 +16,7 @@ import (
"github.com/ollama/ollama/api"
"github.com/ollama/ollama/discover"
"github.com/ollama/ollama/fs/ggml"
"github.com/ollama/ollama/llm"
)
@ -45,8 +46,8 @@ func (mockRunner) Tokenize(_ context.Context, s string) (tokens []int, err error
return
}
func newMockServer(mock *mockRunner) func(discover.GpuInfoList, string, *llm.GGML, []string, []string, api.Options, int) (llm.LlamaServer, error) {
return func(gpus discover.GpuInfoList, model string, ggml *llm.GGML, projectors, system []string, opts api.Options, numParallel int) (llm.LlamaServer, error) {
func newMockServer(mock *mockRunner) func(discover.GpuInfoList, string, *ggml.GGML, []string, []string, api.Options, int) (llm.LlamaServer, error) {
return func(_ discover.GpuInfoList, _ string, _ *ggml.GGML, _, _ []string, _ api.Options, _ int) (llm.LlamaServer, error) {
return mock, nil
}
}
@ -76,7 +77,7 @@ func TestGenerateChat(t *testing.T) {
getGpuFn: discover.GetGPUInfo,
getCpuFn: discover.GetCPUInfo,
reschedDelay: 250 * time.Millisecond,
loadFn: func(req *LlmRequest, ggml *llm.GGML, gpus discover.GpuInfoList, numParallel int) {
loadFn: func(req *LlmRequest, _ *ggml.GGML, _ discover.GpuInfoList, _ int) {
// add small delay to simulate loading
time.Sleep(time.Millisecond)
req.successCh <- &runnerRef{
@ -88,7 +89,7 @@ func TestGenerateChat(t *testing.T) {
go s.sched.Run(context.TODO())
_, digest := createBinFile(t, llm.KV{
_, digest := createBinFile(t, ggml.KV{
"general.architecture": "llama",
"llama.block_count": uint32(1),
"llama.context_length": uint32(8192),
@ -98,7 +99,7 @@ func TestGenerateChat(t *testing.T) {
"tokenizer.ggml.tokens": []string{""},
"tokenizer.ggml.scores": []float32{0},
"tokenizer.ggml.token_type": []int32{0},
}, []llm.Tensor{
}, []ggml.Tensor{
{Name: "token_embd.weight", Shape: []uint64{1}, WriterTo: bytes.NewReader(make([]byte, 4))},
{Name: "blk.0.attn_norm.weight", Shape: []uint64{1}, WriterTo: bytes.NewReader(make([]byte, 4))},
{Name: "blk.0.ffn_down.weight", Shape: []uint64{1}, WriterTo: bytes.NewReader(make([]byte, 4))},
@ -154,10 +155,10 @@ func TestGenerateChat(t *testing.T) {
})
t.Run("missing capabilities chat", func(t *testing.T) {
_, digest := createBinFile(t, llm.KV{
_, digest := createBinFile(t, ggml.KV{
"general.architecture": "bert",
"bert.pooling_type": uint32(0),
}, []llm.Tensor{})
}, []ggml.Tensor{})
w := createRequest(t, s.CreateHandler, api.CreateRequest{
Model: "bert",
Files: map[string]string{"bert.gguf": digest},
@ -612,7 +613,7 @@ func TestGenerate(t *testing.T) {
getGpuFn: discover.GetGPUInfo,
getCpuFn: discover.GetCPUInfo,
reschedDelay: 250 * time.Millisecond,
loadFn: func(req *LlmRequest, ggml *llm.GGML, gpus discover.GpuInfoList, numParallel int) {
loadFn: func(req *LlmRequest, _ *ggml.GGML, _ discover.GpuInfoList, _ int) {
// add small delay to simulate loading
time.Sleep(time.Millisecond)
req.successCh <- &runnerRef{
@ -624,7 +625,7 @@ func TestGenerate(t *testing.T) {
go s.sched.Run(context.TODO())
_, digest := createBinFile(t, llm.KV{
_, digest := createBinFile(t, ggml.KV{
"general.architecture": "llama",
"llama.block_count": uint32(1),
"llama.context_length": uint32(8192),
@ -634,7 +635,7 @@ func TestGenerate(t *testing.T) {
"tokenizer.ggml.tokens": []string{""},
"tokenizer.ggml.scores": []float32{0},
"tokenizer.ggml.token_type": []int32{0},
}, []llm.Tensor{
}, []ggml.Tensor{
{Name: "token_embd.weight", Shape: []uint64{1}, WriterTo: bytes.NewReader(make([]byte, 4))},
{Name: "blk.0.attn_norm.weight", Shape: []uint64{1}, WriterTo: bytes.NewReader(make([]byte, 4))},
{Name: "blk.0.ffn_down.weight", Shape: []uint64{1}, WriterTo: bytes.NewReader(make([]byte, 4))},
@ -686,10 +687,10 @@ func TestGenerate(t *testing.T) {
})
t.Run("missing capabilities generate", func(t *testing.T) {
_, digest := createBinFile(t, llm.KV{
_, digest := createBinFile(t, ggml.KV{
"general.architecture": "bert",
"bert.pooling_type": uint32(0),
}, []llm.Tensor{})
}, []ggml.Tensor{})
w := createRequest(t, s.CreateHandler, api.CreateRequest{
Model: "bert",