refactor convert

server/model.go

@@ -143,30 +143,6 @@ func parseFromZipFile(_ context.Context, file *os.File, digest string, fn func(a
 		return nil, err
 	}
 
-	mf, err := convert.GetModelFormat(tempDir)
-	if err != nil {
-		return nil, err
-	}
-
-	params, err := mf.GetParams(tempDir)
-	if err != nil {
-		return nil, err
-	}
-
-	mArch, err := mf.GetModelArch("", tempDir, params)
-	if err != nil {
-		return nil, err
-	}
-
-	fn(api.ProgressResponse{Status: "processing tensors"})
-	if err := mArch.GetTensors(); err != nil {
-		return nil, err
-	}
-
-	if err := mArch.LoadVocab(); err != nil {
-		return nil, err
-	}
-
 	fn(api.ProgressResponse{Status: "converting model"})
 
 	// TODO(mxyng): this should write directly into a layer
@@ -178,7 +154,7 @@ func parseFromZipFile(_ context.Context, file *os.File, digest string, fn func(a
 	defer temp.Close()
 	defer os.Remove(temp.Name())
 
-	if err = mArch.WriteGGUF(temp); err != nil {
+	if err := convert.Convert(tempDir, temp); err != nil {
 		return nil, err
 	}
 

server/routes_create_test.go

@@ -2,7 +2,6 @@ package server
 
 import (
 	"bytes"
-	"encoding/binary"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -20,7 +19,7 @@ import (
 
 var stream bool = false
 
-func createBinFile(t *testing.T, kv map[string]any, ti []llm.Tensor) string {
+func createBinFile(t *testing.T, kv map[string]any, ti []*llm.Tensor) string {
 	t.Helper()
 
 	f, err := os.CreateTemp(t.TempDir(), "")
@@ -29,7 +28,7 @@ func createBinFile(t *testing.T, kv map[string]any, ti []llm.Tensor) string {
 	}
 	defer f.Close()
 
-	if err := llm.NewGGUFV3(binary.LittleEndian).Encode(f, kv, ti); err != nil {
+	if err := llm.WriteGGUF(f, kv, ti); err != nil {
 		t.Fatal(err)
 	}
 

server/routes_generate_test.go

@@ -101,7 +101,7 @@ func TestGenerateChat(t *testing.T) {
 			"tokenizer.ggml.tokens":         []string{""},
 			"tokenizer.ggml.scores":         []float32{0},
 			"tokenizer.ggml.token_type":     []int32{0},
-		}, []llm.Tensor{
+		}, []*llm.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))},
@@ -149,7 +149,7 @@ func TestGenerateChat(t *testing.T) {
 			Modelfile: fmt.Sprintf("FROM %s", createBinFile(t, llm.KV{
 				"general.architecture": "bert",
 				"bert.pooling_type":    uint32(0),
-			}, []llm.Tensor{})),
+			}, []*llm.Tensor{})),
 			Stream: &stream,
 		})
 
@@ -399,7 +399,7 @@ func TestGenerate(t *testing.T) {
 			"tokenizer.ggml.tokens":         []string{""},
 			"tokenizer.ggml.scores":         []float32{0},
 			"tokenizer.ggml.token_type":     []int32{0},
-		}, []llm.Tensor{
+		}, []*llm.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))},
@@ -447,7 +447,7 @@ func TestGenerate(t *testing.T) {
 			Modelfile: fmt.Sprintf("FROM %s", createBinFile(t, llm.KV{
 				"general.architecture": "bert",
 				"bert.pooling_type":    uint32(0),
-			}, []llm.Tensor{})),
+			}, []*llm.Tensor{})),
 			Stream: &stream,
 		})
 

server/sched_test.go

@@ -3,7 +3,6 @@ package server
 import (
 	"bytes"
 	"context"
-	"encoding/binary"
 	"fmt"
 	"log/slog"
 	"os"
@@ -114,8 +113,7 @@ func newScenarioRequest(t *testing.T, ctx context.Context, modelName string, est
 	require.NoError(t, err)
 	defer f.Close()
 
-	gguf := llm.NewGGUFV3(binary.LittleEndian)
-	err = gguf.Encode(f, llm.KV{
+	require.NoError(t, llm.WriteGGUF(f, llm.KV{
 		"general.architecture":          "llama",
 		"general.name":                  "name",
 		"llama.context_length":          uint32(32),
@@ -126,10 +124,10 @@ func newScenarioRequest(t *testing.T, ctx context.Context, modelName string, est
 		"tokenizer.ggml.tokens":         []string{" "},
 		"tokenizer.ggml.scores":         []float32{0},
 		"tokenizer.ggml.token_type":     []int32{0},
-	}, []llm.Tensor{
+	}, []*llm.Tensor{
 		{Name: "blk.0.attn.weight", Kind: uint32(0), Offset: uint64(0), Shape: []uint64{1, 1, 1, 1}, WriterTo: bytes.NewReader(make([]byte, 32))},
 		{Name: "output.weight", Kind: uint32(0), Offset: uint64(0), Shape: []uint64{1, 1, 1, 1}, WriterTo: bytes.NewReader(make([]byte, 32))},
-	})
+	}))
 	require.NoError(t, err)
 
 	fname := f.Name()