use vanity imports

add stablelm graph calculation

.github/ISSUE_TEMPLATE/10_bug_report.yml

@@ -0,0 +1,60 @@
+name: Bug report
+labels: [bug]
+description: Something isn't working right.
+body:
+  - type: textarea
+    id: description
+    attributes:
+      label: What is the issue?
+      description: What happened? What did you expect to happen?
+    validations:
+      required: true
+  - type: dropdown
+    id: os
+    attributes:
+      label: OS
+      description: Which operating system are you using?
+      multiple: true
+      options:
+        - Linux
+        - macOS
+        - Windows
+        - Docker
+        - WSL2
+    validations:
+      required: false
+  - type: dropdown
+    id: gpu
+    attributes:
+      label: GPU
+      description: Which GPU are you using?
+      multiple: true
+      options:
+        - Nvidia
+        - AMD
+        - Intel
+        - Apple
+        - Other
+    validations:
+      required: false
+  - type: dropdown
+    id: cpu
+    attributes:
+      label: CPU
+      description: Which CPU are you using?
+      multiple: true
+      options:
+        - Intel
+        - AMD
+        - Apple
+        - Other
+    validations:
+      required: false
+  - type: input
+    id: version
+    attributes:
+      label: Ollama version
+      description: What version of Ollama are you using? (`ollama --version`)
+      placeholder: e.g., 0.1.32
+    validations:
+      required: false

.github/ISSUE_TEMPLATE/10_model_request.yml

@@ -1,18 +0,0 @@
-name: Model request
-description: Request a new model for the library
-labels: [mr]
-body:
-  - type: markdown
-    attributes:
-      value: |
-        Please check if your Model request is [already available](https://ollama.com/search) or that you cannot [import it](https://github.com/ollama/ollama/blob/main/docs/import.md#import-a-model) yourself.
-        Tell us about which Model you'd like to see in the library!
-  - type: textarea
-    id: problem
-    attributes:
-      label: What model would you like?
-      description: Please provide a link to the model.
-  - type: markdown
-    attributes:
-      value: |
-        Thanks for filing a model request!

.github/ISSUE_TEMPLATE/20_feature_request.md

@@ -0,0 +1,6 @@
+---
+name: Feature request
+about: Request a new feature
+labels: feature request
+---
+

.github/ISSUE_TEMPLATE/20_feature_request.yml

@@ -1,41 +0,0 @@
-name: Feature request
-description: Propose a new feature
-labels: [needs-triage, fr]
-body:
-  - type: markdown
-    attributes:
-      value: |
-        Please check if your feature request is [already filed](https://github.com/ollama/ollama/issues).
-        Tell us about your idea!
-  - type: textarea
-    id: problem
-    attributes:
-      label: What are you trying to do?
-      description: Tell us about the problem you're trying to solve.
-    validations:
-      required: false
-  - type: textarea
-    id: solution
-    attributes:
-      label: How should we solve this?
-      description: If you have an idea of how you'd like to see this feature work, let us know.
-    validations:
-      required: false
-  - type: textarea
-    id: alternative
-    attributes:
-      label: What is the impact of not solving this?
-      description: (How) Are you currently working around the issue?
-    validations:
-      required: false
-  - type: textarea
-    id: context
-    attributes:
-      label: Anything else?
-      description: Any additional context to share, e.g., links
-    validations:
-      required: false
-  - type: markdown
-    attributes:
-      value: |
-        Thanks for filing a feature request!

.github/ISSUE_TEMPLATE/30_model_request.md

@@ -0,0 +1,5 @@
+---
+name: Model request
+about: Request support for a new model to be added to Ollama
+labels: model request
+---

.github/ISSUE_TEMPLATE/90_bug_report.yml

@@ -1,125 +0,0 @@
-name: Bug report
-description: File a bug report. If you need help, please join our Discord server.
-labels: [needs-triage, bug]
-body:
-  - type: markdown
-    attributes:
-      value: |
-        Please check if your bug is [already filed](https://github.com/ollama/ollama/issues) before filing a new one.
-  - type: textarea
-    id: what-happened
-    attributes:
-      label: What is the issue?
-      description: What happened? What did you expect to happen?
-    validations:
-      required: true
-  - type: textarea
-    id: what-was-expected
-    attributes:
-      label: What did you expect to see?
-      description: What did you expect to see/happen instead?
-    validations:
-      required: false
-  - type: textarea
-    id: steps
-    attributes:
-      label: Steps to reproduce
-      description: What are the steps you took that hit this issue?
-    validations:
-      required: false
-  - type: textarea
-    id: changes
-    attributes:
-      label: Are there any recent changes that introduced the issue?
-      description: If so, what are those changes?
-    validations:
-      required: false
-  - type: dropdown
-    id: os
-    attributes:
-      label: OS
-      description: What OS are you using? You may select more than one.
-      multiple: true
-      options:
-        - Linux
-        - macOS
-        - Windows
-        - Other
-    validations:
-      required: false
-  - type: dropdown
-    id: architecture
-    attributes:
-      label: Architecture
-      description: What architecture are you using? You may select more than one.
-      multiple: true
-      options:
-        - arm64
-        - amd64
-        - x86
-        - Other
-  - type: dropdown
-    id: platform
-    attributes:
-      label: Platform
-      description: What platform are you using? You may select more than one.
-      multiple: true
-      options:
-        - Docker
-        - WSL
-        - WSL2
-    validations:
-      required: false
-  - type: input
-    id: ollama-version
-    attributes:
-      label: Ollama version
-      description: What Ollama version are you using? (`ollama --version`)
-      placeholder: e.g., 1.14.4
-    validations:
-      required: false
-  - type: dropdown
-    id: gpu
-    attributes:
-      label: GPU
-      description: What GPU, if any, are you using? You may select more than one.
-      multiple: true
-      options:
-        - Nvidia
-        - AMD
-        - Intel
-        - Apple
-        - Other
-    validations:
-      required: false
-  - type: textarea
-    id: gpu-info
-    attributes:
-      label: GPU info
-      description: What GPU info do you have? (`nvidia-smi`, `rocminfo`, `system_profiler SPDisplaysDataType`, etc.)
-    validations:
-      required: false
-  - type: dropdown
-    id: cpu
-    attributes:
-      label: CPU
-      description: What CPU are you using? You may select more than one.
-      multiple: true
-      options:
-        - Intel
-        - AMD
-        - Apple
-        - Other
-    validations:
-      required: false
-  - type: textarea
-    id: other-software
-    attributes:
-      label: Other software
-      description: What other software are you using that might be related to this issue?
-    validations:
-      required: false
-  - type: markdown
-    attributes:
-      value: |
-        Thanks for filing a bug report!

README.md

@@ -60,7 +60,6 @@ Here are some example models that can be downloaded:
 | Llama 2 13B        | 13B        | 7.3GB | `ollama run llama2:13b`        |
 | Llama 2 70B        | 70B        | 39GB  | `ollama run llama2:70b`        |
 | Orca Mini          | 3B         | 1.9GB | `ollama run orca-mini`         |
-| Vicuna             | 7B         | 3.8GB | `ollama run vicuna`            |
 | LLaVA              | 7B         | 4.5GB | `ollama run llava`             |
 | Gemma              | 2B         | 1.4GB | `ollama run gemma:2b`          |
 | Gemma              | 7B         | 4.8GB | `ollama run gemma:7b`          |
@@ -378,3 +377,6 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Page Assist](https://github.com/n4ze3m/page-assist) (Chrome Extension)
 - [AI Telegram Bot](https://github.com/tusharhero/aitelegrambot) (Telegram bot using Ollama in backend)
 - [AI ST Completion](https://github.com/yaroslavyaroslav/OpenAI-sublime-text) (Sublime Text 4 AI assistant plugin with Ollama support)
+
+### Supported backends 
+- [llama.cpp](https://github.com/ggerganov/llama.cpp) project founded by Georgi Gerganov. 

api/client.go

@@ -20,8 +20,8 @@ import (
 	"runtime"
 	"strings"
 
-	"github.com/ollama/ollama/format"
-	"github.com/ollama/ollama/version"
+	"ollama.com/format"
+	"ollama.com/version"
 )
 
 // Client encapsulates client state for interacting with the ollama

app/lifecycle/lifecycle.go

@@ -9,8 +9,8 @@ import (
 	"os/signal"
 	"syscall"
 
-	"github.com/ollama/ollama/app/store"
-	"github.com/ollama/ollama/app/tray"
+	"ollama.com/app/store"
+	"ollama.com/app/tray"
 )
 
 func Run() {

app/lifecycle/server.go

@@ -11,7 +11,7 @@ import (
 	"path/filepath"
 	"time"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func getCLIFullPath(command string) string {

app/lifecycle/updater.go

@@ -18,8 +18,8 @@ import (
 	"strings"
 	"time"
 
-	"github.com/ollama/ollama/auth"
-	"github.com/ollama/ollama/version"
+	"ollama.com/auth"
+	"ollama.com/version"
 )
 
 var (

app/main.go

@@ -4,7 +4,7 @@ package main
 // go build -ldflags="-H windowsgui" .
 
 import (
-	"github.com/ollama/ollama/app/lifecycle"
+	"ollama.com/app/lifecycle"
 )
 
 func main() {

app/tray/tray.go

@@ -4,8 +4,8 @@ import (
 	"fmt"
 	"runtime"
 
-	"github.com/ollama/ollama/app/assets"
-	"github.com/ollama/ollama/app/tray/commontray"
+	"ollama.com/app/assets"
+	"ollama.com/app/tray/commontray"
 )
 
 func NewTray() (commontray.OllamaTray, error) {

app/tray/tray_nonwindows.go

@@ -5,7 +5,7 @@ package tray
 import (
 	"fmt"
 
-	"github.com/ollama/ollama/app/tray/commontray"
+	"ollama.com/app/tray/commontray"
 )
 
 func InitPlatformTray(icon, updateIcon []byte) (commontray.OllamaTray, error) {

app/tray/tray_windows.go

@@ -1,8 +1,8 @@
 package tray
 
 import (
-	"github.com/ollama/ollama/app/tray/commontray"
-	"github.com/ollama/ollama/app/tray/wintray"
+	"ollama.com/app/tray/commontray"
+	"ollama.com/app/tray/wintray"
 )
 
 func InitPlatformTray(icon, updateIcon []byte) (commontray.OllamaTray, error) {

app/tray/wintray/tray.go

@@ -13,8 +13,8 @@ import (
 	"sync"
 	"unsafe"
 
-	"github.com/ollama/ollama/app/tray/commontray"
 	"golang.org/x/sys/windows"
+	"ollama.com/app/tray/commontray"
 )
 
 // Helpful sources: https://github.com/golang/exp/blob/master/shiny/driver/internal/win32

cmd/cmd.go

@@ -30,12 +30,12 @@ import (
 	"golang.org/x/exp/slices"
 	"golang.org/x/term"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/format"
-	"github.com/ollama/ollama/parser"
-	"github.com/ollama/ollama/progress"
-	"github.com/ollama/ollama/server"
-	"github.com/ollama/ollama/version"
+	"ollama.com/api"
+	"ollama.com/format"
+	"ollama.com/parser"
+	"ollama.com/progress"
+	"ollama.com/server"
+	"ollama.com/version"
 )
 
 func CreateHandler(cmd *cobra.Command, args []string) error {

cmd/interactive.go

@@ -14,9 +14,9 @@ import (
 	"github.com/spf13/cobra"
 	"golang.org/x/exp/slices"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/progress"
-	"github.com/ollama/ollama/readline"
+	"ollama.com/api"
+	"ollama.com/progress"
+	"ollama.com/readline"
 )
 
 type MultilineState int

cmd/interactive_test.go

@@ -7,7 +7,7 @@ import (
 
 	"github.com/stretchr/testify/assert"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func TestExtractFilenames(t *testing.T) {

cmd/start_darwin.go

@@ -7,7 +7,7 @@ import (
 	"os/exec"
 	"strings"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func startApp(ctx context.Context, client *api.Client) error {

cmd/start_default.go

@@ -6,7 +6,7 @@ import (
 	"context"
 	"fmt"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func startApp(ctx context.Context, client *api.Client) error {

cmd/start_windows.go

@@ -10,7 +10,7 @@ import (
 	"strings"
 	"syscall"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func startApp(ctx context.Context, client *api.Client) error {

convert/convert.go

@@ -13,8 +13,8 @@ import (
 
 	"google.golang.org/protobuf/proto"
 
-	"github.com/ollama/ollama/convert/sentencepiece"
-	"github.com/ollama/ollama/llm"
+	"ollama.com/convert/sentencepiece"
+	"ollama.com/llm"
 )
 
 type Params struct {

convert/gemma.go

@@ -12,7 +12,7 @@ import (
 	"github.com/pdevine/tensor"
 	"github.com/pdevine/tensor/native"
 
-	"github.com/ollama/ollama/llm"
+	"ollama.com/llm"
 )
 
 type GemmaModel struct {

convert/llama.go

@@ -14,7 +14,7 @@ import (
 	"github.com/pdevine/tensor/native"
 	"github.com/x448/float16"
 
-	"github.com/ollama/ollama/llm"
+	"ollama.com/llm"
 )
 
 type LlamaModel struct {

convert/mistral.go

@@ -13,7 +13,7 @@ import (
 	"github.com/pdevine/tensor/native"
 	"github.com/x448/float16"
 
-	"github.com/ollama/ollama/llm"
+	"ollama.com/llm"
 )
 
 type MistralModel struct {

convert/safetensors.go

@@ -16,7 +16,7 @@ import (
 	"github.com/mitchellh/mapstructure"
 	"github.com/x448/float16"
 
-	"github.com/ollama/ollama/llm"
+	"ollama.com/llm"
 )
 
 type safetensorWriterTo struct {

convert/torch.go

@@ -15,7 +15,7 @@ import (
 	"github.com/nlpodyssey/gopickle/types"
 	"github.com/x448/float16"
 
-	"github.com/ollama/ollama/llm"
+	"ollama.com/llm"
 )
 
 type torchWriterTo struct {

examples/go-chat/main.go

@@ -5,7 +5,7 @@ import (
 	"fmt"
 	"log"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func main() {

examples/go-generate-streaming/main.go

@@ -5,7 +5,7 @@ import (
 	"fmt"
 	"log"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func main() {

examples/go-generate/main.go

@@ -5,7 +5,7 @@ import (
 	"fmt"
 	"log"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func main() {

examples/go-multimodal/main.go

@@ -6,7 +6,7 @@ import (
 	"log"
 	"os"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func main() {

examples/go-pull-progress/main.go

@@ -5,7 +5,7 @@ import (
 	"fmt"
 	"log"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func main() {

go.mod

@@ -1,4 +1,4 @@
-module github.com/ollama/ollama
+module ollama.com
 
 go 1.22
 

gpu/gpu.go

@@ -21,7 +21,7 @@ import (
 	"sync"
 	"unsafe"
 
-	"github.com/ollama/ollama/format"
+	"ollama.com/format"
 )
 
 type handles struct {

gpu/gpu_darwin.go

@@ -55,6 +55,6 @@ func getCPUMem() (memInfo, error) {
 	return memInfo{
 		TotalMemory: uint64(C.getPhysicalMemory()),
 		FreeMemory:  0,
-		DeviceCount: 0,
+		DeviceCount: 1,
 	}, nil
 }

integration/basic_test.go

@@ -8,7 +8,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func TestOrcaMiniBlueSky(t *testing.T) {

integration/context_test.go

@@ -8,7 +8,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func TestContextExhaustion(t *testing.T) {

integration/llm_image_test.go

@@ -9,8 +9,8 @@ import (
 	"testing"
 	"time"
 
-	"github.com/ollama/ollama/api"
 	"github.com/stretchr/testify/require"
+	"ollama.com/api"
 )
 
 func TestIntegrationMultimodal(t *testing.T) {

integration/llm_test.go

@@ -9,7 +9,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 // TODO - this would ideally be in the llm package, but that would require some refactoring of interfaces in the server

integration/utils_test.go

@@ -21,9 +21,9 @@ import (
 	"testing"
 	"time"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/app/lifecycle"
 	"github.com/stretchr/testify/assert"
+	"ollama.com/api"
+	"ollama.com/app/lifecycle"
 )
 
 func FindPort() string {

llm/ext_server/server.cpp

@@ -39,6 +39,10 @@
 #include "httplib.h"
 #include "json.hpp"
 
+#if defined(_WIN32)
+#include <windows.h>
+#endif
+
 #include <cstddef>
 #include <thread>
 #include <chrono>
@@ -2770,8 +2774,28 @@ inline void signal_handler(int signal) {
     shutdown_handler(signal);
 }
 
-int main(int argc, char **argv)
-{
+#if defined(_WIN32)
+char* wchar_to_char(const wchar_t* wstr) {
+    if (wstr == nullptr) return nullptr;
+
+    // Determine the number of bytes needed for the UTF-8 string
+    int bytes = WideCharToMultiByte(CP_UTF8, 0, wstr, -1, nullptr, 0, nullptr, nullptr);
+    char* str = new char[bytes];
+
+    // Convert the wide-character string to a UTF-8 string
+    WideCharToMultiByte(CP_UTF8, 0, wstr, -1, str, bytes, nullptr, nullptr);
+    return str;
+}
+
+int wmain(int argc, wchar_t **wargv) {
+    char** argv = new char*[argc];
+    for (int i = 0; i < argc; ++i) {
+        argv[i] = wchar_to_char(wargv[i]);
+    }
+#else
+int main(int argc, char **argv) {
+#endif
+
 #if SERVER_VERBOSE != 1
     log_disable();
 #endif
@@ -3282,6 +3306,11 @@ int main(int argc, char **argv)
         return (ctrl_type == CTRL_C_EVENT) ? (signal_handler(SIGINT), true) : false;
     };
     SetConsoleCtrlHandler(reinterpret_cast<PHANDLER_ROUTINE>(console_ctrl_handler), true);
+
+    for (int i = 0; i < argc; ++i) {
+        delete[] argv[i];
+    }
+    delete[] argv;
 #endif
     llama.queue_tasks.start_loop();
     svr.stop();

llm/ggml.go

@@ -164,7 +164,8 @@ func (ts Tensors) Layers() map[string]Layer {
 	for _, t := range ts {
 		parts := strings.Split(t.Name, ".")
 		if parts[0] == "blk" {
-			parts = parts[1:]
+			// join first and second part, e.g. blk.%d
+			parts = append([]string{fmt.Sprintf("%s.%s", parts[0], parts[1])}, parts[2:]...)
 		}
 
 		if _, ok := layers[parts[0]]; !ok {
@@ -380,6 +381,12 @@ func (llm GGML) GraphSize(context, batch uint64) (partialOffload, fullOffload ui
 		)
 
 		partialOffload = 4*batch*(2*embedding+vocab) + embedding*vocab*105/128
+	case "stablelm":
+		fullOffload = 4 * batch * (context*(1+heads) + 3*embedding + 2)
+		partialOffload = max(
+			4*batch*(vocab+2*embedding),
+			fullOffload,
+		)
 	}
 
 	return

llm/gguf.go

@@ -248,13 +248,17 @@ func (llm *gguf) Decode(rs io.ReadSeeker) error {
 	}
 
 	padding := llm.padding(offset, int64(alignment))
-	if _, err := rs.Seek(padding-offset, io.SeekCurrent); err != nil {
+	if _, err := rs.Seek(padding, io.SeekCurrent); err != nil {
 		return err
 	}
 
 	for _, tensor := range llm.tensors {
-		padded := (int64(tensor.size()) + int64(alignment) - 1) & ^(int64(alignment) - 1)
-		if _, err := rs.Seek(padded, io.SeekCurrent); err != nil {
+		if _, err := rs.Seek(int64(tensor.size()), io.SeekCurrent); err != nil {
+			return err
+		}
+
+		padding := llm.padding(int64(tensor.size()), int64(alignment))
+		if _, err := rs.Seek(padding, io.SeekCurrent); err != nil {
 			return err
 		}
 	}
@@ -623,8 +627,9 @@ func (llm *gguf) Encode(ws io.WriteSeeker, kv KV, tensors []Tensor) error {
 		return err
 	}
 
-	padding := llm.padding(offset, 32)
-	if err := binary.Write(ws, llm.ByteOrder, bytes.Repeat([]byte{0}, int(padding-offset))); err != nil {
+	var alignment int64 = 32
+	padding := llm.padding(offset, alignment)
+	if err := binary.Write(ws, llm.ByteOrder, bytes.Repeat([]byte{0}, int(padding))); err != nil {
 		return err
 	}
 
@@ -638,8 +643,8 @@ func (llm *gguf) Encode(ws io.WriteSeeker, kv KV, tensors []Tensor) error {
 			return err
 		}
 
-		padding := llm.padding(offset, 32)
-		if err := binary.Write(ws, llm.ByteOrder, bytes.Repeat([]byte{0}, int(padding-offset))); err != nil {
+		padding := llm.padding(offset, alignment)
+		if err := binary.Write(ws, llm.ByteOrder, bytes.Repeat([]byte{0}, int(padding))); err != nil {
 			return err
 		}
 	}
@@ -648,5 +653,5 @@ func (llm *gguf) Encode(ws io.WriteSeeker, kv KV, tensors []Tensor) error {
 }
 
 func (gguf) padding(offset, align int64) int64 {
-	return (offset + align - 1) / align * align
+	return (align - offset%align) % align
 }

llm/payload.go

@@ -14,7 +14,7 @@ import (
 	"golang.org/x/exp/slices"
 	"golang.org/x/sync/errgroup"
 
-	"github.com/ollama/ollama/gpu"
+	"ollama.com/gpu"
 )
 
 var errPayloadMissing = fmt.Errorf("expected payloads not included in this build of ollama")

llm/server.go

@@ -21,9 +21,9 @@ import (
 	"strings"
 	"time"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/format"
-	"github.com/ollama/ollama/gpu"
+	"ollama.com/api"
+	"ollama.com/format"
+	"ollama.com/gpu"
 )
 
 // LlamaServer is an instance of the llama.cpp server
@@ -79,6 +79,9 @@ func NewLlamaServer(model string, adapters, projectors []string, opts api.Option
 		graphFullOffload = graphPartialOffload
 	}
 
+	graphFullOffload *= uint64(info.DeviceCount)
+	graphPartialOffload *= uint64(info.DeviceCount)
+
 	// memoryRequiredTotal represents the memory required for full GPU offloading (all layers)
 	memoryRequiredTotal := memoryMinimum + graphFullOffload
 
@@ -94,7 +97,7 @@ func NewLlamaServer(model string, adapters, projectors []string, opts api.Option
 	var layerCount int
 	layers := ggml.Tensors().Layers()
 	for i := 0; i < int(ggml.KV().BlockCount()); i++ {
-		memoryLayer := layers[fmt.Sprintf("%d", i)].size()
+		memoryLayer := layers[fmt.Sprintf("blk.%d", i)].size()
 
 		// KV is proportional to the number of layers
 		memoryLayer += kv / ggml.KV().BlockCount()
@@ -106,7 +109,13 @@ func NewLlamaServer(model string, adapters, projectors []string, opts api.Option
 		}
 	}
 
-	memoryLayerOutput := layers["output"].size()
+	var memoryLayerOutput uint64
+	for k, v := range layers {
+		if !strings.HasPrefix(k, "blk.") {
+			memoryLayerOutput += v.size()
+		}
+	}
+
 	memoryRequiredTotal += memoryLayerOutput
 
 	if info.Library == "metal" && memoryRequiredTotal > info.TotalMemory {
@@ -121,16 +130,47 @@ func NewLlamaServer(model string, adapters, projectors []string, opts api.Option
 		opts.NumGPU = layerCount
 	}
 
+	memoryWeights := memoryRequiredTotal - memoryMinimum - graphFullOffload - kv
+
 	slog.Info(
 		"offload to gpu",
-		"reallayers", opts.NumGPU,
-		"layers", layerCount,
-		"required", format.HumanBytes2(memoryRequiredTotal),
-		"used", format.HumanBytes2(memoryRequiredPartial),
-		"available", format.HumanBytes2(memoryAvailable),
-		"kv", format.HumanBytes2(kv),
-		"fulloffload", format.HumanBytes2(graphFullOffload),
-		"partialoffload", format.HumanBytes2(graphPartialOffload),
+		slog.Group(
+			"layers",
+			// actual number of layers offloaded
+			"real", opts.NumGPU,
+			// estimated number of layers that can be offloaded
+			"estimate", layerCount,
+		),
+		slog.Group(
+			"memory",
+			// memory available for offloading
+			"available", format.HumanBytes2(memoryAvailable),
+			slog.Group(
+				"required",
+				// memory required for full offloading
+				"full", format.HumanBytes2(memoryRequiredTotal),
+				// memory required to offload layers.estimate layers
+				"partial", format.HumanBytes2(memoryRequiredPartial),
+				// memory of KV cache
+				"kv", format.HumanBytes2(kv),
+			),
+			slog.Group(
+				"weights",
+				// memory of the weights
+				"total", format.HumanBytes2(memoryWeights),
+				// memory of repeating layers
+				"repeating", format.HumanBytes2(memoryWeights-memoryLayerOutput),
+				// memory of non-repeating layers
+				"nonrepeating", format.HumanBytes2(memoryLayerOutput),
+			),
+			slog.Group(
+				"graph",
+				// memory of graph when fully offloaded
+				"full", format.HumanBytes2(graphFullOffload),
+				// memory of graph when not fully offloaded
+				"partial", format.HumanBytes2(graphPartialOffload),
+			),
+		),
 	)
 
 	if len(adapters) > 1 {

main.go

@@ -3,8 +3,8 @@ package main
 import (
 	"context"
 
-	"github.com/ollama/ollama/cmd"
 	"github.com/spf13/cobra"
+	"ollama.com/cmd"
 )
 
 func main() {

openai/openai.go

@@ -11,7 +11,7 @@ import (
 	"time"
 
 	"github.com/gin-gonic/gin"
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 type Error struct {

progress/bar.go

@@ -6,8 +6,8 @@ import (
 	"strings"
 	"time"
 
-	"github.com/ollama/ollama/format"
 	"golang.org/x/term"
+	"ollama.com/format"
 )
 
 type Bar struct {

server/auth.go

@@ -15,8 +15,8 @@ import (
 	"strings"
 	"time"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/auth"
+	"ollama.com/api"
+	"ollama.com/auth"
 )
 
 type registryChallenge struct {

server/download.go

@@ -21,8 +21,8 @@ import (
 
 	"golang.org/x/sync/errgroup"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/format"
+	"ollama.com/api"
+	"ollama.com/format"
 )
 
 const maxRetries = 6

server/images.go

@@ -24,12 +24,12 @@ import (
 
 	"golang.org/x/exp/slices"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/convert"
-	"github.com/ollama/ollama/format"
-	"github.com/ollama/ollama/llm"
-	"github.com/ollama/ollama/parser"
-	"github.com/ollama/ollama/version"
+	"ollama.com/api"
+	"ollama.com/convert"
+	"ollama.com/format"
+	"ollama.com/llm"
+	"ollama.com/parser"
+	"ollama.com/version"
 )
 
 type registryOptions struct {

server/prompt.go

@@ -7,7 +7,7 @@ import (
 	"text/template"
 	"text/template/parse"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 // isResponseNode checks if the node contains .Response

server/prompt_test.go

@@ -4,7 +4,7 @@ import (
 	"strings"
 	"testing"
 
-	"github.com/ollama/ollama/api"
+	"ollama.com/api"
 )
 
 func TestPrompt(t *testing.T) {

server/routes.go

@@ -27,12 +27,12 @@ import (
 	"github.com/gin-gonic/gin"
 	"golang.org/x/exp/slices"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/gpu"
-	"github.com/ollama/ollama/llm"
-	"github.com/ollama/ollama/openai"
-	"github.com/ollama/ollama/parser"
-	"github.com/ollama/ollama/version"
+	"ollama.com/api"
+	"ollama.com/gpu"
+	"ollama.com/llm"
+	"ollama.com/openai"
+	"ollama.com/parser"
+	"ollama.com/version"
 )
 
 var mode string = gin.DebugMode

server/routes_test.go

@@ -16,9 +16,9 @@ import (
 
 	"github.com/stretchr/testify/assert"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/parser"
-	"github.com/ollama/ollama/version"
+	"ollama.com/api"
+	"ollama.com/parser"
+	"ollama.com/version"
 )
 
 func Test_Routes(t *testing.T) {

server/upload.go

@@ -16,9 +16,9 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/format"
 	"golang.org/x/sync/errgroup"
+	"ollama.com/api"
+	"ollama.com/format"
 )
 
 var blobUploadManager sync.Map

types/model/name.go

@@ -12,7 +12,7 @@ import (
 	"strings"
 	"sync"
 
-	"github.com/ollama/ollama/types/structs"
+	"ollama.com/types/structs"
 )
 
 // Errors
@@ -185,8 +185,8 @@ func parseMask(s string) Name {
 	return r
 }
 
-func MustParseName(s, defaults string) Name {
-	r := ParseName(s, "")
+func MustParseName(s, fill string) Name {
+	r := ParseName(s, fill)
 	if !r.IsValid() {
 		panic("invalid Name: " + s)
 	}
@@ -521,6 +521,8 @@ func parts(s string) iter_Seq2[PartKind, string] {
 						return
 					}
 					state, j, partLen = PartModel, i, 0
+				case PartHost:
+					// noop: support for host:port
 				default:
 					yield(PartExtraneous, s[i+1:j])
 					return
@@ -643,6 +645,15 @@ func (r Name) Filepath() string {
 	return filepath.Join(r.parts[:]...)
 }
 
+// FilepathNoBuild returns a complete, canonicalized, relative file path using
+// the parts of a complete Name, but without the build part.
+func (r Name) FilepathNoBuild() string {
+	for i := range PartBuild {
+		r.parts[i] = strings.ToLower(r.parts[i])
+	}
+	return filepath.Join(r.parts[:PartBuild]...)
+}
+
 // isValidPart reports if s contains all valid characters for the given
 // part kind.
 func isValidPart(kind PartKind, s string) bool {
@@ -669,6 +680,9 @@ func isValidByteFor(kind PartKind, c byte) bool {
 	if kind == PartNamespace && c == '.' {
 		return false
 	}
+	if kind == PartHost && c == ':' {
+		return true
+	}
 	if c == '.' || c == '-' {
 		return true
 	}

types/model/name_test.go

@@ -40,6 +40,7 @@ var testNames = map[string]fields{
 	"user/model":                     {namespace: "user", model: "model"},
 	"example.com/ns/mistral:7b+Q4_0": {host: "example.com", namespace: "ns", model: "mistral", tag: "7b", build: "Q4_0"},
 	"example.com/ns/mistral:7b+X":    {host: "example.com", namespace: "ns", model: "mistral", tag: "7b", build: "X"},
+	"localhost:5000/ns/mistral":      {host: "localhost:5000", namespace: "ns", model: "mistral"},
 
 	// invalid digest
 	"mistral:latest@invalid256-": {},
@@ -318,6 +319,13 @@ func TestNameGoString(t *testing.T) {
 	}
 }
 
+func TestDisplayLongest(t *testing.T) {
+	g := ParseName("example.com/library/mistral:latest+Q4_0", FillNothing).DisplayLongest()
+	if g != "example.com/library/mistral:latest" {
+		t.Errorf("got = %q; want %q", g, "example.com/library/mistral:latest")
+	}
+}
+
 func TestDisplayShortest(t *testing.T) {
 	cases := []struct {
 		in        string
@@ -478,30 +486,36 @@ func TestNamePath(t *testing.T) {
 	}
 }
 
-func TestNameFromFilepath(t *testing.T) {
+func TestNameFilepath(t *testing.T) {
 	cases := []struct {
-		in   string
-		want string
+		in          string
+		want        string
+		wantNoBuild string
 	}{
 		{
-			in:   "example.com/library/mistral:latest+Q4_0",
-			want: "example.com/library/mistral/latest/Q4_0",
+			in:          "example.com/library/mistral:latest+Q4_0",
+			want:        "example.com/library/mistral/latest/Q4_0",
+			wantNoBuild: "example.com/library/mistral/latest",
 		},
 		{
-			in:   "Example.Com/Library/Mistral:Latest+Q4_0",
-			want: "example.com/library/mistral/latest/Q4_0",
+			in:          "Example.Com/Library/Mistral:Latest+Q4_0",
+			want:        "example.com/library/mistral/latest/Q4_0",
+			wantNoBuild: "example.com/library/mistral/latest",
 		},
 		{
-			in:   "Example.Com/Library/Mistral:Latest+Q4_0",
-			want: "example.com/library/mistral/latest/Q4_0",
+			in:          "Example.Com/Library/Mistral:Latest+Q4_0",
+			want:        "example.com/library/mistral/latest/Q4_0",
+			wantNoBuild: "example.com/library/mistral/latest",
 		},
 		{
-			in:   "example.com/library/mistral:latest",
-			want: "example.com/library/mistral/latest",
+			in:          "example.com/library/mistral:latest",
+			want:        "example.com/library/mistral/latest",
+			wantNoBuild: "example.com/library/mistral/latest",
 		},
 		{
-			in:   "",
-			want: "",
+			in:          "",
+			want:        "",
+			wantNoBuild: "",
 		},
 	}
 	for _, tt := range cases {
@@ -513,6 +527,11 @@ func TestNameFromFilepath(t *testing.T) {
 			if g != tt.want {
 				t.Errorf("got = %q; want %q", g, tt.want)
 			}
+			g = p.FilepathNoBuild()
+			g = filepath.ToSlash(g)
+			if g != tt.wantNoBuild {
+				t.Errorf("got = %q; want %q", g, tt.wantNoBuild)
+			}
 		})
 	}
 }