ml/backend/ggml: stable sort devices by score (#9081)

removing the channel tag from the url so it will always go to the current stable channel.

CMakeLists.txt

@@ -104,6 +104,10 @@ if(CMAKE_HIP_COMPILER)
     if(AMDGPU_TARGETS)
         add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/ml/backend/ggml/ggml/src/ggml-hip)
 
+        if (WIN32)
+            target_compile_definitions(ggml-hip PRIVATE GGML_CUDA_NO_PEER_COPY=1)
+        endif()
+
         set(OLLAMA_HIP_INSTALL_DIR ${OLLAMA_INSTALL_DIR}/rocm)
         install(TARGETS ggml-hip
             RUNTIME_DEPENDENCIES

README.md

@@ -437,9 +437,10 @@ See the [API documentation](./docs/api.md) for all endpoints.
 
 - [Pacman](https://archlinux.org/packages/extra/x86_64/ollama/)
 - [Gentoo](https://github.com/gentoo/guru/tree/master/app-misc/ollama)
+- [Homebrew](https://formulae.brew.sh/formula/ollama)
 - [Helm Chart](https://artifacthub.io/packages/helm/ollama-helm/ollama)
 - [Guix channel](https://codeberg.org/tusharhero/ollama-guix)
-- [Nix package](https://search.nixos.org/packages?channel=24.05&show=ollama&from=0&size=50&sort=relevance&type=packages&query=ollama)
+- [Nix package](https://search.nixos.org/packages?show=ollama&from=0&size=50&sort=relevance&type=packages&query=ollama)
 - [Flox](https://flox.dev/blog/ollama-part-one)
 
 ### Libraries
@@ -494,7 +495,7 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [multi-llm-ts](https://github.com/nbonamy/multi-llm-ts) (A Typescript/JavaScript library allowing access to different LLM in unified API)
 - [LlmTornado](https://github.com/lofcz/llmtornado) (C# library providing a unified interface for major FOSS & Commercial inference APIs)
 - [Ollama for Zig](https://github.com/dravenk/ollama-zig)
-- [Abso](https://github.com/lunary-ai/abso/blob/main/README.md#ollama) (OpenAI-compatible TypeScript SDK for any LLM provider)
+- [Abso](https://github.com/lunary-ai/abso) (OpenAI-compatible TypeScript SDK for any LLM provider)
 
 ### Mobile
 

api/client.go

@@ -18,6 +18,7 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"net/http"
@@ -29,28 +30,6 @@ import (
 	"github.com/ollama/ollama/version"
 )
 
-// StatusError is an error with an HTTP status code and message,
-// it is parsed on the client-side and not returned from the API
-type StatusError struct {
-	StatusCode int    // e.g. 200
-	Status     string // e.g. "200 OK"
-	ErrorResponse
-}
-
-func (e StatusError) Error() string {
-	switch {
-	case e.Status != "" && e.Err != "":
-		return fmt.Sprintf("%s: %s", e.Status, e.Err)
-	case e.Status != "":
-		return e.Status
-	case e.Err != "":
-		return e.Err
-	default:
-		// this should not happen
-		return "something went wrong, please see the ollama server logs for details"
-	}
-}
-
 // Client encapsulates client state for interacting with the ollama
 // service. Use [ClientFromEnvironment] to create new Clients.
 type Client struct {
@@ -68,7 +47,7 @@ func checkError(resp *http.Response, body []byte) error {
 	err := json.Unmarshal(body, &apiError)
 	if err != nil {
 		// Use the full body as the message if we fail to decode a response.
-		apiError.Err = string(body)
+		apiError.ErrorMessage = string(body)
 	}
 
 	return apiError
@@ -153,7 +132,7 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData
 const maxBufferSize = 512 * format.KiloByte
 
 func (c *Client) stream(ctx context.Context, method, path string, data any, fn func([]byte) error) error {
-	var buf io.Reader
+	var buf *bytes.Buffer
 	if data != nil {
 		bts, err := json.Marshal(data)
 		if err != nil {
@@ -184,22 +163,24 @@ func (c *Client) stream(ctx context.Context, method, path string, data any, fn f
 	scanBuf := make([]byte, 0, maxBufferSize)
 	scanner.Buffer(scanBuf, maxBufferSize)
 	for scanner.Scan() {
-		bts := scanner.Bytes()
+		var errorResponse struct {
+			Error string `json:"error,omitempty"`
+		}
 
-		var errorResponse ErrorResponse
+		bts := scanner.Bytes()
 		if err := json.Unmarshal(bts, &errorResponse); err != nil {
 			return fmt.Errorf("unmarshal: %w", err)
 		}
 
-		if errorResponse.Err != "" {
-			return errorResponse
+		if errorResponse.Error != "" {
+			return errors.New(errorResponse.Error)
 		}
 
 		if response.StatusCode >= http.StatusBadRequest {
 			return StatusError{
-				StatusCode:    response.StatusCode,
-				Status:        response.Status,
-				ErrorResponse: errorResponse,
+				StatusCode:   response.StatusCode,
+				Status:       response.Status,
+				ErrorMessage: errorResponse.Error,
 			}
 		}
 

api/client_test.go

@@ -1,12 +1,6 @@
 package api
 
 import (
-	"context"
-	"encoding/json"
-	"fmt"
-	"net/http"
-	"net/http/httptest"
-	"net/url"
 	"testing"
 )
 
@@ -49,270 +43,3 @@ func TestClientFromEnvironment(t *testing.T) {
 		})
 	}
 }
-
-// testError represents an internal error type for testing different error formats
-type testError struct {
-	message    string         // basic error message
-	structured *ErrorResponse // structured error response, nil for basic format
-	statusCode int
-}
-
-func (e testError) Error() string {
-	return e.message
-}
-
-func TestClientStream(t *testing.T) {
-	testCases := []struct {
-		name      string
-		responses []any
-		wantErr   string
-	}{
-		{
-			name: "basic error format",
-			responses: []any{
-				testError{
-					message:    "test error message",
-					statusCode: http.StatusBadRequest,
-				},
-			},
-			wantErr: "test error message",
-		},
-		{
-			name: "structured error format",
-			responses: []any{
-				testError{
-					message: "test structured error",
-					structured: &ErrorResponse{
-						Err:  "test structured error",
-						Hint: "test hint",
-					},
-					statusCode: http.StatusBadRequest,
-				},
-			},
-			wantErr: "test structured error\ntest hint",
-		},
-		{
-			name: "error after chunks - basic format",
-			responses: []any{
-				ChatResponse{Message: Message{Content: "partial 1"}},
-				ChatResponse{Message: Message{Content: "partial 2"}},
-				testError{
-					message:    "mid-stream basic error",
-					statusCode: http.StatusOK,
-				},
-			},
-			wantErr: "mid-stream basic error",
-		},
-		{
-			name: "error after chunks - structured format",
-			responses: []any{
-				ChatResponse{Message: Message{Content: "partial 1"}},
-				ChatResponse{Message: Message{Content: "partial 2"}},
-				testError{
-					message: "mid-stream structured error",
-					structured: &ErrorResponse{
-						Err:  "mid-stream structured error",
-						Hint: "additional context",
-					},
-					statusCode: http.StatusOK,
-				},
-			},
-			wantErr: "mid-stream structured error\nadditional context",
-		},
-		{
-			name: "successful stream completion",
-			responses: []any{
-				ChatResponse{Message: Message{Content: "chunk 1"}},
-				ChatResponse{Message: Message{Content: "chunk 2"}},
-				ChatResponse{
-					Message:    Message{Content: "final chunk"},
-					Done:       true,
-					DoneReason: "stop",
-				},
-			},
-		},
-	}
-
-	for _, tc := range testCases {
-		t.Run(tc.name, func(t *testing.T) {
-			ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-				flusher, ok := w.(http.Flusher)
-				if !ok {
-					t.Fatal("expected http.Flusher")
-				}
-
-				w.Header().Set("Content-Type", "application/x-ndjson")
-
-				for _, resp := range tc.responses {
-					if errResp, ok := resp.(testError); ok {
-						w.WriteHeader(errResp.statusCode)
-						var err error
-						if errResp.structured != nil {
-							err = json.NewEncoder(w).Encode(errResp.structured)
-						} else {
-							err = json.NewEncoder(w).Encode(map[string]string{
-								"error": errResp.message,
-							})
-						}
-						if err != nil {
-							t.Fatal("failed to encode error response:", err)
-						}
-						return
-					}
-
-					if err := json.NewEncoder(w).Encode(resp); err != nil {
-						t.Fatalf("failed to encode response: %v", err)
-					}
-					flusher.Flush()
-				}
-			}))
-			defer ts.Close()
-
-			client := NewClient(&url.URL{Scheme: "http", Host: ts.Listener.Addr().String()}, http.DefaultClient)
-
-			var receivedChunks []ChatResponse
-			err := client.stream(context.Background(), http.MethodPost, "/v1/chat", nil, func(chunk []byte) error {
-				var resp ChatResponse
-				if err := json.Unmarshal(chunk, &resp); err != nil {
-					return fmt.Errorf("failed to unmarshal chunk: %w", err)
-				}
-				receivedChunks = append(receivedChunks, resp)
-				return nil
-			})
-
-			if tc.wantErr != "" {
-				if err == nil {
-					t.Fatalf("got nil, want error %q", tc.wantErr)
-				}
-				if err.Error() != tc.wantErr {
-					t.Errorf("error message mismatch: got %q, want %q", err.Error(), tc.wantErr)
-				}
-			} else {
-				if err != nil {
-					t.Errorf("got error %q, want nil", err)
-				}
-			}
-		})
-	}
-}
-
-func TestClientDo(t *testing.T) {
-	testCases := []struct {
-		name     string
-		response any
-		wantErr  string
-	}{
-		{
-			name: "basic error format",
-			response: testError{
-				message:    "test error message",
-				statusCode: http.StatusBadRequest,
-			},
-			wantErr: "test error message",
-		},
-		{
-			name: "structured error format",
-			response: testError{
-				message: "test structured error",
-				structured: &ErrorResponse{
-					Err:  "test structured error",
-					Hint: "test hint",
-				},
-				statusCode: http.StatusBadRequest,
-			},
-			wantErr: "test structured error",
-		},
-		{
-			name: "server error - basic format",
-			response: testError{
-				message:    "internal error",
-				statusCode: http.StatusInternalServerError,
-			},
-			wantErr: "internal error",
-		},
-		{
-			name: "server error - structured format",
-			response: testError{
-				message: "internal server error",
-				structured: &ErrorResponse{
-					Err:  "internal server error",
-					Hint: "please try again later",
-				},
-				statusCode: http.StatusInternalServerError,
-			},
-			wantErr: "internal server error",
-		},
-		{
-			name: "successful response",
-			response: struct {
-				ID      string `json:"id"`
-				Success bool   `json:"success"`
-			}{
-				ID:      "msg_123",
-				Success: true,
-			},
-		},
-	}
-
-	for _, tc := range testCases {
-		t.Run(tc.name, func(t *testing.T) {
-			ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-				if errResp, ok := tc.response.(testError); ok {
-					w.WriteHeader(errResp.statusCode)
-					var err error
-					if errResp.structured != nil {
-						err = json.NewEncoder(w).Encode(errResp.structured)
-					} else {
-						err = json.NewEncoder(w).Encode(map[string]string{
-							"error": errResp.message,
-						})
-					}
-					if err != nil {
-						t.Fatal("failed to encode error response:", err)
-					}
-					return
-				}
-
-				w.Header().Set("Content-Type", "application/json")
-				if err := json.NewEncoder(w).Encode(tc.response); err != nil {
-					t.Fatalf("failed to encode response: %v", err)
-				}
-			}))
-			defer ts.Close()
-
-			client := NewClient(&url.URL{Scheme: "http", Host: ts.Listener.Addr().String()}, http.DefaultClient)
-
-			var resp struct {
-				ID      string `json:"id"`
-				Success bool   `json:"success"`
-			}
-			err := client.do(context.Background(), http.MethodPost, "/v1/messages", nil, &resp)
-
-			if tc.wantErr != "" {
-				if err == nil {
-					t.Fatalf("got nil, want error %q", tc.wantErr)
-				}
-				if err.Error() != tc.wantErr {
-					t.Errorf("error message mismatch: got %q, want %q", err.Error(), tc.wantErr)
-				}
-				return
-			}
-
-			if err != nil {
-				t.Errorf("got error %q, want nil", err)
-			}
-
-			if expectedResp, ok := tc.response.(struct {
-				ID      string `json:"id"`
-				Success bool   `json:"success"`
-			}); ok {
-				if resp.ID != expectedResp.ID {
-					t.Errorf("response ID mismatch: got %q, want %q", resp.ID, expectedResp.ID)
-				}
-				if resp.Success != expectedResp.Success {
-					t.Errorf("response Success mismatch: got %v, want %v", resp.Success, expectedResp.Success)
-				}
-			}
-		})
-	}
-}

api/types.go

@@ -12,6 +12,27 @@ import (
 	"time"
 )
 
+// StatusError is an error with an HTTP status code and message.
+type StatusError struct {
+	StatusCode   int
+	Status       string
+	ErrorMessage string `json:"error"`
+}
+
+func (e StatusError) Error() string {
+	switch {
+	case e.Status != "" && e.ErrorMessage != "":
+		return fmt.Sprintf("%s: %s", e.Status, e.ErrorMessage)
+	case e.Status != "":
+		return e.Status
+	case e.ErrorMessage != "":
+		return e.ErrorMessage
+	default:
+		// this should not happen
+		return "something went wrong, please see the ollama server logs for details"
+	}
+}
+
 // ImageData represents the raw binary data of an image file.
 type ImageData []byte
 
@@ -640,22 +661,6 @@ func (d *Duration) UnmarshalJSON(b []byte) (err error) {
 	return nil
 }
 
-// ErrorResponse implements a structured error interface that is returned from the Ollama server
-type ErrorResponse struct {
-	// Err is the error from the server. It helps with debugging the code-path
-	Err string `json:"error"`
-
-	// Hint is a user-friendly message about what went wrong, with suggested troubleshooting
-	Hint string `json:"hint"`
-}
-
-func (e ErrorResponse) Error() string {
-	if e.Hint == "" {
-		return e.Err
-	}
-	return fmt.Sprintf("%s\n%s", e.Err, e.Hint)
-}
-
 // FormatParams converts specified parameter options to their correct types
 func FormatParams(params map[string][]string) (map[string]interface{}, error) {
 	opts := Options{}

llama/patches/0014-sort-devices-by-score.patch

@@ -8,7 +8,7 @@ Subject: [PATCH] sort devices by score
  1 file changed, 13 insertions(+), 8 deletions(-)
 
 diff --git a/ggml/src/ggml-backend-reg.cpp b/ggml/src/ggml-backend-reg.cpp
-index 899d16f2..ac5cda07 100644
+index 899d16f2..135f7df0 100644
 --- a/ggml/src/ggml-backend-reg.cpp
 +++ b/ggml/src/ggml-backend-reg.cpp
 @@ -150,7 +150,7 @@ struct ggml_backend_reg_entry {
@@ -29,7 +29,7 @@ index 899d16f2..ac5cda07 100644
          if (!reg) {
              return;
          }
-@@ -206,15 +206,15 @@ struct ggml_backend_registry {
+@@ -206,15 +206,20 @@ struct ggml_backend_registry {
  #endif
          backends.push_back({ reg, std::move(handle) });
          for (size_t i = 0; i < ggml_backend_reg_dev_count(reg); i++) {
@@ -45,10 +45,15 @@ index 899d16f2..ac5cda07 100644
  #endif
 -        devices.push_back(device);
 +        devices.push_back({device, score});
++        std::stable_sort(devices.begin(), devices.end(),
++            [](const auto & a, const auto & b) {
++                return a.second > b.second;
++            }
++        );
      }
  
      ggml_backend_reg_t load_backend(const std::wstring & path, bool silent) {
-@@ -257,7 +257,7 @@ struct ggml_backend_registry {
+@@ -257,7 +262,7 @@ struct ggml_backend_registry {
  
          GGML_LOG_INFO("%s: loaded %s backend from %s\n", __func__, ggml_backend_reg_name(reg), utf16_to_utf8(path).c_str());
  
@@ -57,7 +62,7 @@ index 899d16f2..ac5cda07 100644
  
          return reg;
      }
-@@ -280,7 +280,7 @@ struct ggml_backend_registry {
+@@ -280,7 +285,7 @@ struct ggml_backend_registry {
          // remove devices
          devices.erase(
              std::remove_if(devices.begin(), devices.end(),
@@ -66,17 +71,12 @@ index 899d16f2..ac5cda07 100644
              devices.end());
  
          // remove backend
-@@ -338,7 +338,12 @@ size_t ggml_backend_dev_count() {
+@@ -338,7 +343,7 @@ size_t ggml_backend_dev_count() {
  
  ggml_backend_dev_t ggml_backend_dev_get(size_t index) {
      GGML_ASSERT(index < ggml_backend_dev_count());
 -    return get_reg().devices[index];
-+    auto devices = get_reg().devices;
-+    if (!std::is_heap(devices.begin(), devices.end())) {
-+        std::make_heap(devices.begin(), devices.end(), [](const auto & a, const auto & b) { return a.second < b.second; });
-+    }
-+
-+    return devices[index].first;
++    return get_reg().devices[index].first;
  }
  
  ggml_backend_dev_t ggml_backend_dev_by_name(const char * name) {

llama/patches/0017-try-catch-backend-load.patch

@@ -0,0 +1,69 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Michael Yang <mxyng@pm.me>
+Date: Tue, 11 Feb 2025 14:06:36 -0800
+Subject: [PATCH] try/catch backend load
+
+---
+ ggml/src/ggml-backend-reg.cpp | 45 ++++++++++++++++++-----------------
+ 1 file changed, 23 insertions(+), 22 deletions(-)
+
+diff --git a/ggml/src/ggml-backend-reg.cpp b/ggml/src/ggml-backend-reg.cpp
+index 135f7df0..84b21dd8 100644
+--- a/ggml/src/ggml-backend-reg.cpp
++++ b/ggml/src/ggml-backend-reg.cpp
+@@ -512,32 +512,33 @@ static ggml_backend_reg_t ggml_backend_load_best(const char * name, bool silent,
+         }
+         fs::directory_iterator dir_it(search_path, fs::directory_options::skip_permission_denied);
+         for (const auto & entry : dir_it) {
+-            if (entry.is_regular_file()) {
+-                std::wstring filename = entry.path().filename().wstring();
+-                std::wstring ext = entry.path().extension().wstring();
+-                if (filename.find(file_prefix) == 0 && ext == backend_filename_suffix()) {
+-                    dl_handle_ptr handle { dl_load_library(entry.path().wstring()) };
+-                    if (!handle && !silent) {
+-                        GGML_LOG_ERROR("%s: failed to load %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
+-                    }
+-                    if (handle) {
++            try {
++                if (entry.is_regular_file()) {
++                    std::wstring filename = entry.path().filename().wstring();
++                    std::wstring ext = entry.path().extension().wstring();
++                    if (filename.find(file_prefix) == 0 && ext == backend_filename_suffix()) {
++                        dl_handle_ptr handle { dl_load_library(entry.path().wstring()) };
++                        if (!handle) {
++                            GGML_LOG_ERROR("%s: failed to load %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
++                            continue;
++                        }
++
+                         auto score_fn = (ggml_backend_score_t) dl_get_sym(handle.get(), "ggml_backend_score");
+-                        if (score_fn) {
+-                            int s = score_fn();
+-#ifndef NDEBUG
+-                            GGML_LOG_DEBUG("%s: %s score: %d\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str(), s);
+-#endif
+-                            if (s > best_score) {
+-                                best_score = s;
+-                                best_path = entry.path().wstring();
+-                            }
+-                        } else {
+-                            if (!silent) {
+-                                GGML_LOG_INFO("%s: failed to find ggml_backend_score in %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
+-                            }
++                        if (!score_fn) {
++                            GGML_LOG_DEBUG("%s: failed to find ggml_backend_score in %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
++                            continue;
++                        }
++
++                        int s = score_fn();
++                        GGML_LOG_DEBUG("%s: %s score: %d\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str(), s);
++                        if (s > best_score) {
++                            best_score = s;
++                            best_path = entry.path().wstring();
+                         }
+                     }
+                 }
++            } catch (const std::exception & e) {
++                GGML_LOG_ERROR("%s: failed to load %s: %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str(), e.what());
+             }
+         }
+     }

ml/backend/ggml/ggml/src/ggml-backend-reg.cpp

@@ -215,6 +215,11 @@ struct ggml_backend_registry {
         GGML_LOG_DEBUG("%s: registered device %s (%s)\n", __func__, ggml_backend_dev_name(device), ggml_backend_dev_description(device));
 #endif
         devices.push_back({device, score});
+        std::stable_sort(devices.begin(), devices.end(),
+            [](const auto & a, const auto & b) {
+                return a.second > b.second;
+            }
+        );
     }
 
     ggml_backend_reg_t load_backend(const std::wstring & path, bool silent) {
@@ -338,12 +343,7 @@ size_t ggml_backend_dev_count() {
 
 ggml_backend_dev_t ggml_backend_dev_get(size_t index) {
     GGML_ASSERT(index < ggml_backend_dev_count());
-    auto devices = get_reg().devices;
-    if (!std::is_heap(devices.begin(), devices.end())) {
-        std::make_heap(devices.begin(), devices.end(), [](const auto & a, const auto & b) { return a.second < b.second; });
-    }
-
-    return devices[index].first;
+    return get_reg().devices[index].first;
 }
 
 ggml_backend_dev_t ggml_backend_dev_by_name(const char * name) {
@@ -512,32 +512,33 @@ static ggml_backend_reg_t ggml_backend_load_best(const char * name, bool silent,
         }
         fs::directory_iterator dir_it(search_path, fs::directory_options::skip_permission_denied);
         for (const auto & entry : dir_it) {
-            if (entry.is_regular_file()) {
-                std::wstring filename = entry.path().filename().wstring();
-                std::wstring ext = entry.path().extension().wstring();
-                if (filename.find(file_prefix) == 0 && ext == backend_filename_suffix()) {
-                    dl_handle_ptr handle { dl_load_library(entry.path().wstring()) };
-                    if (!handle && !silent) {
-                        GGML_LOG_ERROR("%s: failed to load %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
-                    }
-                    if (handle) {
+            try {
+                if (entry.is_regular_file()) {
+                    std::wstring filename = entry.path().filename().wstring();
+                    std::wstring ext = entry.path().extension().wstring();
+                    if (filename.find(file_prefix) == 0 && ext == backend_filename_suffix()) {
+                        dl_handle_ptr handle { dl_load_library(entry.path().wstring()) };
+                        if (!handle) {
+                            GGML_LOG_ERROR("%s: failed to load %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
+                            continue;
+                        }
+
                         auto score_fn = (ggml_backend_score_t) dl_get_sym(handle.get(), "ggml_backend_score");
-                        if (score_fn) {
-                            int s = score_fn();
-#ifndef NDEBUG
-                            GGML_LOG_DEBUG("%s: %s score: %d\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str(), s);
-#endif
-                            if (s > best_score) {
-                                best_score = s;
-                                best_path = entry.path().wstring();
-                            }
-                        } else {
-                            if (!silent) {
-                                GGML_LOG_INFO("%s: failed to find ggml_backend_score in %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
-                            }
+                        if (!score_fn) {
+                            GGML_LOG_DEBUG("%s: failed to find ggml_backend_score in %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str());
+                            continue;
+                        }
+
+                        int s = score_fn();
+                        GGML_LOG_DEBUG("%s: %s score: %d\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str(), s);
+                        if (s > best_score) {
+                            best_score = s;
+                            best_path = entry.path().wstring();
                         }
                     }
                 }
+            } catch (const std::exception & e) {
+                GGML_LOG_ERROR("%s: failed to load %s: %s\n", __func__, utf16_to_utf8(entry.path().wstring()).c_str(), e.what());
             }
         }
     }

ml/backend/ggml/ggml/src/ggml.go

@@ -79,6 +79,11 @@ var OnceLoad = sync.OnceFunc(func() {
 			continue
 		}
 
+		if abspath != filepath.Dir(exe) && !strings.Contains(abspath, filepath.FromSlash("lib/ollama")) {
+			slog.Debug("skipping path which is not part of ollama", "path", abspath)
+			continue
+		}
+
 		if _, ok := visited[abspath]; !ok {
 			func() {
 				slog.Debug("ggml backend load all from path", "path", abspath)

openai/openai.go

@@ -610,14 +610,14 @@ type EmbedWriter struct {
 }
 
 func (w *BaseWriter) writeError(data []byte) (int, error) {
-	var er api.ErrorResponse // error response is used here to parse the error message
-	err := json.Unmarshal(data, &er)
+	var serr api.StatusError
+	err := json.Unmarshal(data, &serr)
 	if err != nil {
 		return 0, err
 	}
 
 	w.ResponseWriter.Header().Set("Content-Type", "application/json")
-	err = json.NewEncoder(w.ResponseWriter).Encode(NewError(http.StatusInternalServerError, er.Err))
+	err = json.NewEncoder(w.ResponseWriter).Encode(NewError(http.StatusInternalServerError, serr.Error()))
 	if err != nil {
 		return 0, err
 	}

server/images.go

@@ -550,7 +550,7 @@ func PullModel(ctx context.Context, name string, regOpts *registryOptions, fn fu
 
 	manifest, err = pullModelManifest(ctx, mp, regOpts)
 	if err != nil {
-		return fmt.Errorf("pull model manifest: %w", err)
+		return fmt.Errorf("pull model manifest: %s", err)
 	}
 
 	var layers []Layer
@@ -629,18 +629,13 @@ func PullModel(ctx context.Context, name string, regOpts *registryOptions, fn fu
 	return nil
 }
 
-var ErrRemoteModelNotFound = errors.New("model not found")
-
 func pullModelManifest(ctx context.Context, mp ModelPath, regOpts *registryOptions) (*Manifest, error) {
 	requestURL := mp.BaseURL().JoinPath("v2", mp.GetNamespaceRepository(), "manifests", mp.Tag)
 
 	headers := make(http.Header)
 	headers.Set("Accept", "application/vnd.docker.distribution.manifest.v2+json")
 	resp, err := makeRequestWithRetry(ctx, http.MethodGet, requestURL, headers, nil, regOpts)
-	if errors.Is(err, os.ErrNotExist) {
-		// The model was not found on the remote registry
-		return nil, fmt.Errorf("%w: %s", ErrRemoteModelNotFound, err)
-	} else if err != nil {
+	if err != nil {
 		return nil, err
 	}
 	defer resp.Body.Close()

server/routes.go

@@ -564,8 +564,7 @@ func (s *Server) PullHandler(c *gin.Context) {
 		return
 	}
 
-	reqName := cmp.Or(req.Model, req.Name)
-	name := model.ParseName(reqName)
+	name := model.ParseName(cmp.Or(req.Model, req.Name))
 	if !name.IsValid() {
 		c.AbortWithStatusJSON(http.StatusBadRequest, gin.H{"error": errtypes.InvalidModelNameErrMsg})
 		return
@@ -592,18 +591,7 @@ func (s *Server) PullHandler(c *gin.Context) {
 		defer cancel()
 
 		if err := PullModel(ctx, name.DisplayShortest(), regOpts, fn); err != nil {
-			if errors.Is(err, ErrRemoteModelNotFound) {
-				hint := fmt.Sprintf("Model %q not found - please check the model name is correct and try again", reqName)
-				if name.Host == DefaultRegistry {
-					hint = fmt.Sprintf("Model %q not found - search available models at: https://ollama.com/search?q=%s", reqName, reqName)
-				}
-				ch <- api.ErrorResponse{
-					Err:  err.Error(),
-					Hint: hint,
-				}
-			} else {
-				ch <- gin.H{"error": err.Error()}
-			}
+			ch <- gin.H{"error": err.Error()}
 		}
 	}()