mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-21 22:33:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'ollama:main' into main

Browse Source
This commit is contained in:
likelovewant
2025-10-21 19:38:31 +08:00
committed by GitHub
parent cb13784a11 5fe7ba1b9b
commit 7f551c41e7
13 changed files with 137 additions and 75 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
convert/convert_test.go
View File

@@ -18,6 +18,7 @@ import (
"strings" "strings"
"testing" "testing"
"github.com/google/go-cmp/cmp"
"github.com/ollama/ollama/fs/ggml" "github.com/ollama/ollama/fs/ggml"
) )
@@ -339,13 +340,8 @@ func TestConvertAdapter(t *testing.T) {
} }
actual := generateResultsJSON(t, r, m.KV(), m.Tensors()) actual := generateResultsJSON(t, r, m.KV(), m.Tensors())
if diff := cmp.Diff(c.Expected, actual); diff != "" {
for _, k := range slices.Sorted(maps.Keys(c.Expected)) { t.Errorf("mismatch (-want +got):\n%s", diff)
if v, ok := actual[k]; !ok {
t.Errorf("missing %s", k)
} else if v != c.Expected[k] {
t.Errorf("unexpected %s: want %s, got %s", k, c.Expected[k], v)
}
} }
}) })
} }

8
discover/runner.go
View File

@@ -88,6 +88,7 @@ func GPUDevices(ctx context.Context, runners []FilteredRunnerDiscovery) []ml.Dev
// times concurrently leading to memory contention // times concurrently leading to memory contention
// TODO refactor so we group the lib dirs and do serial per version, but parallel for different libs // TODO refactor so we group the lib dirs and do serial per version, but parallel for different libs
for dir := range libDirs { for dir := range libDirs {
bootstrapTimeout := 30 * time.Second
var dirs []string var dirs []string
if dir != "" { if dir != "" {
if requested != "" && filepath.Base(dir) != requested { if requested != "" && filepath.Base(dir) != requested {
@@ -102,11 +103,16 @@ func GPUDevices(ctx context.Context, runners []FilteredRunnerDiscovery) []ml.Dev
} else { } else {
dirs = []string{LibOllamaPath, dir} dirs = []string{LibOllamaPath, dir}
} }
// ROCm can take a long time on some systems, so give it more time before giving up
if dir != "" && strings.Contains(filepath.Base(dir), "rocm") {
bootstrapTimeout = 60 * time.Second
}
// Typically bootstrapping takes < 1s, but on some systems, with devices // Typically bootstrapping takes < 1s, but on some systems, with devices
// in low power/idle mode, initialization can take multiple seconds. We // in low power/idle mode, initialization can take multiple seconds. We
// set a long timeout just for bootstrap discovery to reduce the chance // set a long timeout just for bootstrap discovery to reduce the chance
// of giving up too quickly // of giving up too quickly
ctx1stPass, cancel := context.WithTimeout(ctx, 30*time.Second) ctx1stPass, cancel := context.WithTimeout(ctx, bootstrapTimeout)
defer cancel() defer cancel()
// For this pass, we retain duplicates in case any are incompatible with some libraries // For this pass, we retain duplicates in case any are incompatible with some libraries

18
fs/ggml/gguf.go
View File

@@ -509,7 +509,10 @@ func writeGGUFArray[S ~[]E, E any](w io.Writer, t uint32, s S) error {
} }
func WriteGGUF(f *os.File, kv KV, ts []*Tensor) error { func WriteGGUF(f *os.File, kv KV, ts []*Tensor) error {
alignment := kv.Uint("general.alignment", 32) arch := kv.String("general.architecture")
if arch == "" {
return fmt.Errorf("architecture not set")
}
if err := binary.Write(f, binary.LittleEndian, []byte("GGUF")); err != nil { if err := binary.Write(f, binary.LittleEndian, []byte("GGUF")); err != nil {
return err return err
@@ -528,7 +531,7 @@ func WriteGGUF(f *os.File, kv KV, ts []*Tensor) error {
} }
for _, key := range slices.Sorted(maps.Keys(kv)) { for _, key := range slices.Sorted(maps.Keys(kv)) {
if err := ggufWriteKV(f, key, kv[key]); err != nil { if err := ggufWriteKV(f, arch, key, kv[key]); err != nil {
return err return err
} }
} }
@@ -543,6 +546,8 @@ func WriteGGUF(f *os.File, kv KV, ts []*Tensor) error {
}, },
) )
alignment := kv.Uint("general.alignment", 32)
var s uint64 var s uint64
for i := range ts { for i := range ts {
ts[i].Offset = s ts[i].Offset = s
@@ -574,7 +579,14 @@ func WriteGGUF(f *os.File, kv KV, ts []*Tensor) error {
return g.Wait() return g.Wait()
} }
func ggufWriteKV(ws io.WriteSeeker, k string, v any) error { func ggufWriteKV(ws io.WriteSeeker, arch, k string, v any) error {
if !strings.HasPrefix(k, arch+".") &&
!strings.HasPrefix(k, "general.") &&
!strings.HasPrefix(k, "adapter.") &&
!strings.HasPrefix(k, "tokenizer.") {
k = arch + "." + k
}
slog.Debug(k, "type", fmt.Sprintf("%T", v)) slog.Debug(k, "type", fmt.Sprintf("%T", v))
if err := binary.Write(ws, binary.LittleEndian, uint64(len(k))); err != nil { if err := binary.Write(ws, binary.LittleEndian, uint64(len(k))); err != nil {
return err return err

14
fs/ggml/gguf_test.go
View File

@@ -39,7 +39,12 @@ func TestWriteGGUF(t *testing.T) {
defer w.Close() defer w.Close()
if err := WriteGGUF(w, KV{ if err := WriteGGUF(w, KV{
"general.alignment": uint32(16), "general.architecture": "test",
"general.alignment": uint32(16),
"test.key": "value",
"attention.key": "value2",
"tokenizer.key": "value3",
"adapter.key": "value4",
}, ts); err != nil { }, ts); err != nil {
t.Fatal(err) t.Fatal(err)
} }
@@ -56,14 +61,19 @@ func TestWriteGGUF(t *testing.T) {
} }
if diff := cmp.Diff(KV{ if diff := cmp.Diff(KV{
"general.architecture": "test",
"general.alignment": uint32(16), "general.alignment": uint32(16),
"general.parameter_count": uint64(54), "general.parameter_count": uint64(54),
"test.key": "value",
"test.attention.key": "value2",
"tokenizer.key": "value3",
"adapter.key": "value4",
}, ff.KV()); diff != "" { }, ff.KV()); diff != "" {
t.Errorf("Mismatch (-want +got):\n%s", diff) t.Errorf("Mismatch (-want +got):\n%s", diff)
} }
if diff := cmp.Diff(Tensors{ if diff := cmp.Diff(Tensors{
Offset: 592, Offset: 800,
items: []*Tensor{ items: []*Tensor{
{Name: "blk.0.attn_k.weight", Offset: 0, Shape: []uint64{2, 3}}, {Name: "blk.0.attn_k.weight", Offset: 0, Shape: []uint64{2, 3}},
{Name: "blk.0.attn_norm.weight", Offset: 32, Shape: []uint64{2, 3}}, {Name: "blk.0.attn_norm.weight", Offset: 32, Shape: []uint64{2, 3}},

13
integration/embed_test.go
View File

@@ -258,6 +258,19 @@ func TestAllMiniLMEmbedTruncate(t *testing.T) {
} }
}, },
}, },
{
name: "boundary truncation",
request: api.EmbedRequest{
Model: "all-minilm",
Input: "why is the sky blue? Why is the sky blue? hi there my",
Options: map[string]any{"num_ctx": 16},
},
check: func(res *api.EmbedResponse, err error) {
if err != nil {
t.Fatal(err)
}
},
},
} }
for _, req := range cases { for _, req := range cases {

30
llama/patches/0026-GPU-discovery-enhancements.patch
View File

@@ -6,20 +6,20 @@ Subject: [PATCH] GPU discovery enhancements
Expose more information about the devices through backend props, and leverage Expose more information about the devices through backend props, and leverage
management libraries for more accurate VRAM usage reporting if available. management libraries for more accurate VRAM usage reporting if available.
--- ---
ggml/include/ggml-backend.h | 9 + ggml/include/ggml-backend.h | 11 +
ggml/src/CMakeLists.txt | 2 + ggml/src/CMakeLists.txt | 2 +
ggml/src/ggml-cuda/ggml-cuda.cu | 72 +++++ ggml/src/ggml-cuda/ggml-cuda.cu | 74 +++++
ggml/src/ggml-cuda/vendors/hip.h | 3 + ggml/src/ggml-cuda/vendors/hip.h | 3 +
ggml/src/ggml-impl.h | 8 + ggml/src/ggml-impl.h | 8 +
ggml/src/ggml-metal/ggml-metal.cpp | 2 + ggml/src/ggml-metal/ggml-metal.cpp | 2 +
ggml/src/mem_hip.cpp | 449 +++++++++++++++++++++++++++++ ggml/src/mem_hip.cpp | 449 +++++++++++++++++++++++++++++
ggml/src/mem_nvml.cpp | 209 ++++++++++++++ ggml/src/mem_nvml.cpp | 209 ++++++++++++++
8 files changed, 754 insertions(+) 8 files changed, 758 insertions(+)
create mode 100644 ggml/src/mem_hip.cpp create mode 100644 ggml/src/mem_hip.cpp
create mode 100644 ggml/src/mem_nvml.cpp create mode 100644 ggml/src/mem_nvml.cpp
diff --git a/ggml/include/ggml-backend.h b/ggml/include/ggml-backend.h diff --git a/ggml/include/ggml-backend.h b/ggml/include/ggml-backend.h
index ba181d09..09ff75f9 100644 index ba181d09d..094fc3c82 100644
--- a/ggml/include/ggml-backend.h --- a/ggml/include/ggml-backend.h
+++ b/ggml/include/ggml-backend.h +++ b/ggml/include/ggml-backend.h
@@ -169,6 +169,17 @@ extern "C" { @@ -169,6 +169,17 @@ extern "C" {
@@ -41,7 +41,7 @@ index ba181d09..09ff75f9 100644
GGML_API const char * ggml_backend_dev_name(ggml_backend_dev_t device); GGML_API const char * ggml_backend_dev_name(ggml_backend_dev_t device);
diff --git a/ggml/src/CMakeLists.txt b/ggml/src/CMakeLists.txt diff --git a/ggml/src/CMakeLists.txt b/ggml/src/CMakeLists.txt
index 0609c650..aefe43bd 100644 index 0609c6503..aefe43bdd 100644
--- a/ggml/src/CMakeLists.txt --- a/ggml/src/CMakeLists.txt
+++ b/ggml/src/CMakeLists.txt +++ b/ggml/src/CMakeLists.txt
@@ -209,6 +209,8 @@ add_library(ggml-base @@ -209,6 +209,8 @@ add_library(ggml-base
@@ -54,7 +54,7 @@ index 0609c650..aefe43bd 100644
target_include_directories(ggml-base PRIVATE .) target_include_directories(ggml-base PRIVATE .)
diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
index 87c6c34a..6a278b5e 100644 index 87c6c34a4..816597d2f 100644
--- a/ggml/src/ggml-cuda/ggml-cuda.cu --- a/ggml/src/ggml-cuda/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda/ggml-cuda.cu +++ b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -261,6 +261,16 @@ static ggml_cuda_device_info ggml_cuda_init() { @@ -261,6 +261,16 @@ static ggml_cuda_device_info ggml_cuda_init() {
@@ -161,21 +161,23 @@ index 87c6c34a..6a278b5e 100644
bool host_buffer = getenv("GGML_CUDA_NO_PINNED") == nullptr; bool host_buffer = getenv("GGML_CUDA_NO_PINNED") == nullptr;
#ifdef GGML_CUDA_NO_PEER_COPY #ifdef GGML_CUDA_NO_PEER_COPY
bool events = false; bool events = false;
@@ -4087,6 +4149,8 @@ ggml_backend_reg_t ggml_backend_cuda_reg() { @@ -4087,6 +4149,7 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
std::lock_guard<std::mutex> lock(mutex); std::lock_guard<std::mutex> lock(mutex);
if (!initialized) { if (!initialized) {
ggml_backend_cuda_reg_context * ctx = new ggml_backend_cuda_reg_context; ggml_backend_cuda_reg_context * ctx = new ggml_backend_cuda_reg_context;
+ int driverVersion = 0; + int driverVersion = 0;
+ CUDA_CHECK(cudaDriverGetVersion(&driverVersion));
for (int i = 0; i < ggml_cuda_info().device_count; i++) { for (int i = 0; i < ggml_cuda_info().device_count; i++) {
ggml_backend_cuda_device_context * dev_ctx = new ggml_backend_cuda_device_context; ggml_backend_cuda_device_context * dev_ctx = new ggml_backend_cuda_device_context;
@@ -4102,6 +4166,14 @@ ggml_backend_reg_t ggml_backend_cuda_reg() { @@ -4102,6 +4165,17 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
snprintf(pci_bus_id, sizeof(pci_bus_id), "%04x:%02x:%02x.0", prop.pciDomainID, prop.pciBusID, prop.pciDeviceID); snprintf(pci_bus_id, sizeof(pci_bus_id), "%04x:%02x:%02x.0", prop.pciDomainID, prop.pciBusID, prop.pciDeviceID);
dev_ctx->pci_bus_id = pci_bus_id; dev_ctx->pci_bus_id = pci_bus_id;
+ dev_ctx->major = prop.major; + dev_ctx->major = prop.major;
+ dev_ctx->minor = prop.minor; + dev_ctx->minor = prop.minor;
+ if (driverVersion == 0) {
+ CUDA_CHECK(cudaDriverGetVersion(&driverVersion));
+ }
+ dev_ctx->driver_major = driverVersion / 1000; + dev_ctx->driver_major = driverVersion / 1000;
+ dev_ctx->driver_minor = (driverVersion - (dev_ctx->driver_major * 1000)) / 10; + dev_ctx->driver_minor = (driverVersion - (dev_ctx->driver_major * 1000)) / 10;
+ dev_ctx->integrated = prop.integrated; + dev_ctx->integrated = prop.integrated;
@@ -186,7 +188,7 @@ index 87c6c34a..6a278b5e 100644
/* .iface = */ ggml_backend_cuda_device_interface, /* .iface = */ ggml_backend_cuda_device_interface,
/* .reg = */ &reg, /* .reg = */ &reg,
diff --git a/ggml/src/ggml-cuda/vendors/hip.h b/ggml/src/ggml-cuda/vendors/hip.h diff --git a/ggml/src/ggml-cuda/vendors/hip.h b/ggml/src/ggml-cuda/vendors/hip.h
index 1f06be80..2f9ef2dc 100644 index 1f06be80e..2f9ef2dc0 100644
--- a/ggml/src/ggml-cuda/vendors/hip.h --- a/ggml/src/ggml-cuda/vendors/hip.h
+++ b/ggml/src/ggml-cuda/vendors/hip.h +++ b/ggml/src/ggml-cuda/vendors/hip.h
@@ -5,6 +5,8 @@ @@ -5,6 +5,8 @@
@@ -207,7 +209,7 @@ index 1f06be80..2f9ef2dc 100644
#define cudaErrorPeerAccessAlreadyEnabled hipErrorPeerAccessAlreadyEnabled #define cudaErrorPeerAccessAlreadyEnabled hipErrorPeerAccessAlreadyEnabled
#define cudaErrorPeerAccessNotEnabled hipErrorPeerAccessNotEnabled #define cudaErrorPeerAccessNotEnabled hipErrorPeerAccessNotEnabled
diff --git a/ggml/src/ggml-impl.h b/ggml/src/ggml-impl.h diff --git a/ggml/src/ggml-impl.h b/ggml/src/ggml-impl.h
index d0fb3bcc..80597b6e 100644 index d0fb3bcca..80597b6ea 100644
--- a/ggml/src/ggml-impl.h --- a/ggml/src/ggml-impl.h
+++ b/ggml/src/ggml-impl.h +++ b/ggml/src/ggml-impl.h
@@ -638,6 +638,14 @@ static inline bool ggml_can_fuse(const struct ggml_cgraph * cgraph, int node_idx @@ -638,6 +638,14 @@ static inline bool ggml_can_fuse(const struct ggml_cgraph * cgraph, int node_idx
@@ -226,7 +228,7 @@ index d0fb3bcc..80597b6e 100644
} }
#endif #endif
diff --git a/ggml/src/ggml-metal/ggml-metal.cpp b/ggml/src/ggml-metal/ggml-metal.cpp diff --git a/ggml/src/ggml-metal/ggml-metal.cpp b/ggml/src/ggml-metal/ggml-metal.cpp
index f2ff9f32..f356e4a0 100644 index f2ff9f322..f356e4a0a 100644
--- a/ggml/src/ggml-metal/ggml-metal.cpp --- a/ggml/src/ggml-metal/ggml-metal.cpp
+++ b/ggml/src/ggml-metal/ggml-metal.cpp +++ b/ggml/src/ggml-metal/ggml-metal.cpp
@@ -535,6 +535,7 @@ static enum ggml_backend_dev_type ggml_backend_metal_device_get_type(ggml_backen @@ -535,6 +535,7 @@ static enum ggml_backend_dev_type ggml_backend_metal_device_get_type(ggml_backen
@@ -247,7 +249,7 @@ index f2ff9f32..f356e4a0 100644
/* .host_buffer = */ false, /* .host_buffer = */ false,
diff --git a/ggml/src/mem_hip.cpp b/ggml/src/mem_hip.cpp diff --git a/ggml/src/mem_hip.cpp b/ggml/src/mem_hip.cpp
new file mode 100644 new file mode 100644
index 00000000..8ef19b8c index 000000000..8ef19b8cf
--- /dev/null --- /dev/null
+++ b/ggml/src/mem_hip.cpp +++ b/ggml/src/mem_hip.cpp
@@ -0,0 +1,449 @@ @@ -0,0 +1,449 @@
@@ -703,7 +705,7 @@ index 00000000..8ef19b8c
\ No newline at end of file \ No newline at end of file
diff --git a/ggml/src/mem_nvml.cpp b/ggml/src/mem_nvml.cpp diff --git a/ggml/src/mem_nvml.cpp b/ggml/src/mem_nvml.cpp
new file mode 100644 new file mode 100644
index 00000000..c9073cef index 000000000..c9073cef0
--- /dev/null --- /dev/null
+++ b/ggml/src/mem_nvml.cpp +++ b/ggml/src/mem_nvml.cpp
@@ -0,0 +1,209 @@ @@ -0,0 +1,209 @@

4
ml/backend/ggml/ggml/src/ggml-cuda/ggml-cuda.cu vendored
View File

@@ -4159,7 +4159,6 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
if (!initialized) { if (!initialized) {
ggml_backend_cuda_reg_context * ctx = new ggml_backend_cuda_reg_context; ggml_backend_cuda_reg_context * ctx = new ggml_backend_cuda_reg_context;
int driverVersion = 0; int driverVersion = 0;
CUDA_CHECK(cudaDriverGetVersion(&driverVersion));
for (int i = 0; i < ggml_cuda_info().device_count; i++) { for (int i = 0; i < ggml_cuda_info().device_count; i++) {
ggml_backend_cuda_device_context * dev_ctx = new ggml_backend_cuda_device_context; ggml_backend_cuda_device_context * dev_ctx = new ggml_backend_cuda_device_context;
@@ -4177,6 +4176,9 @@ ggml_backend_reg_t ggml_backend_cuda_reg() {
dev_ctx->major = prop.major; dev_ctx->major = prop.major;
dev_ctx->minor = prop.minor; dev_ctx->minor = prop.minor;
if (driverVersion == 0) {
CUDA_CHECK(cudaDriverGetVersion(&driverVersion));
}
dev_ctx->driver_major = driverVersion / 1000; dev_ctx->driver_major = driverVersion / 1000;
dev_ctx->driver_minor = (driverVersion - (dev_ctx->driver_major * 1000)) / 10; dev_ctx->driver_minor = (driverVersion - (dev_ctx->driver_major * 1000)) / 10;
dev_ctx->integrated = prop.integrated; dev_ctx->integrated = prop.integrated;

3
model/parsers/qwen3vl.go
View File

@@ -181,9 +181,6 @@ func (p *Qwen3VLParser) eat() ([]qwenEvent, bool) {
split := strings.SplitN(p.buffer.String(), thinkingCloseTag, 2) split := strings.SplitN(p.buffer.String(), thinkingCloseTag, 2)
// before := split[0] // before := split[0]
before := strings.TrimRightFunc(split[0], unicode.IsSpace) before := strings.TrimRightFunc(split[0], unicode.IsSpace)
if len(before) == 0 {
slog.Warn("qwen tool call closing tag found but no content before it")
}
after := strings.TrimLeftFunc(split[1], unicode.IsSpace) after := strings.TrimLeftFunc(split[1], unicode.IsSpace)
if len(before) > 0 { if len(before) > 0 {
events = append(events, qwenEventThinkingContent{content: before}) events = append(events, qwenEventThinkingContent{content: before})

6
parser/parser_test.go
View File

@@ -7,6 +7,7 @@ import (
"errors" "errors"
"fmt" "fmt"
"io" "io"
"maps"
"os" "os"
"strings" "strings"
"testing" "testing"
@@ -799,7 +800,10 @@ func createBinFile(t *testing.T, kv map[string]any, ti []*ggml.Tensor) (string,
} }
defer f.Close() defer f.Close()
if err := ggml.WriteGGUF(f, kv, ti); err != nil { base := map[string]any{"general.architecture": "test"}
maps.Copy(base, kv)
if err := ggml.WriteGGUF(f, base, ti); err != nil {
t.Fatal(err) t.Fatal(err)
} }
// Calculate sha256 of file // Calculate sha256 of file

9
runner/llamarunner/runner.go
View File

@@ -697,7 +697,14 @@ func (s *Server) embeddings(w http.ResponseWriter, r *http.Request) {
w.Header().Set("Content-Type", "application/json") w.Header().Set("Content-Type", "application/json")
seq, err := s.NewSequence(req.Content, nil, NewSequenceParams{embedding: true}) seq, err := s.NewSequence(req.Content, nil, NewSequenceParams{
embedding: true,
// TODO (jmorganca): this should be provided by the server via the
// request options and truncated here in the runner, instead of relying on
// the server's truncate logic
truncate: true,
})
if err != nil { if err != nil {
http.Error(w, fmt.Sprintf("Failed to create new sequence: %v", err), http.StatusInternalServerError) http.Error(w, fmt.Sprintf("Failed to create new sequence: %v", err), http.StatusInternalServerError)
return return

9
runner/ollamarunner/runner.go
View File

@@ -946,7 +946,14 @@ func (s *Server) embeddings(w http.ResponseWriter, r *http.Request) {
} }
w.Header().Set("Content-Type", "application/json") w.Header().Set("Content-Type", "application/json")
seq, err := s.NewSequence(req.Content, nil, NewSequenceParams{embedding: true}) seq, err := s.NewSequence(req.Content, nil, NewSequenceParams{
embedding: true,
// TODO (jmorganca): this should be provided by the server via the
// request options and truncated here in the runner, instead of relying on
// the server's truncate logic
truncate: true,
})
if err != nil { if err != nil {
http.Error(w, fmt.Sprintf("failed to create new sequence: %v", err), http.StatusInternalServerError) http.Error(w, fmt.Sprintf("failed to create new sequence: %v", err), http.StatusInternalServerError)
return return

78
server/routes_create_test.go
View File

@@ -7,6 +7,7 @@ import (
"encoding/json" "encoding/json"
"fmt" "fmt"
"io" "io"
"maps"
"net/http" "net/http"
"net/http/httptest" "net/http/httptest"
"os" "os"
@@ -17,6 +18,8 @@ import (
"testing" "testing"
"github.com/gin-gonic/gin" "github.com/gin-gonic/gin"
gocmp "github.com/google/go-cmp/cmp"
gocmpopts "github.com/google/go-cmp/cmp/cmpopts"
"github.com/ollama/ollama/api" "github.com/ollama/ollama/api"
"github.com/ollama/ollama/envconfig" "github.com/ollama/ollama/envconfig"
@@ -38,7 +41,10 @@ func createBinFile(t *testing.T, kv map[string]any, ti []*ggml.Tensor) (string,
} }
defer f.Close() defer f.Close()
if err := ggml.WriteGGUF(f, kv, ti); err != nil { base := map[string]any{"general.architecture": "test"}
maps.Copy(base, kv)
if err := ggml.WriteGGUF(f, base, ti); err != nil {
t.Fatal(err) t.Fatal(err)
} }
// Calculate sha256 of file // Calculate sha256 of file
@@ -102,8 +108,8 @@ func checkFileExists(t *testing.T, p string, expect []string) {
t.Fatal(err) t.Fatal(err)
} }
if !slices.Equal(actual, expect) { if diff := gocmp.Diff(expect, actual, gocmpopts.SortSlices(strings.Compare), gocmpopts.EquateEmpty()); diff != "" {
t.Fatalf("expected slices to be equal %v", actual) t.Errorf("file exists mismatch (-want +got):\n%s", diff)
} }
} }
@@ -133,8 +139,8 @@ func TestCreateFromBin(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-6bcdb8859d417753645538d7bbfbd7ca91a3f0c191aef5379c53c05e86b669dd"),
filepath.Join(p, "blobs", "sha256-ca239d7bd8ea90e4a5d2e6bf88f8d74a47b14336e73eb4e18bed4dd325018116"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
}) })
} }
@@ -177,8 +183,8 @@ func TestCreateFromModel(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-6bcdb8859d417753645538d7bbfbd7ca91a3f0c191aef5379c53c05e86b669dd"),
filepath.Join(p, "blobs", "sha256-ca239d7bd8ea90e4a5d2e6bf88f8d74a47b14336e73eb4e18bed4dd325018116"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
}) })
} }
@@ -206,9 +212,9 @@ func TestCreateRemovesLayers(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-b507b9c2f6ca642bffcd06665ea7c91f235fd32daeefdf875a0f938db05fb315"), filepath.Join(p, "blobs", "sha256-b507b9c2f6ca642bffcd06665ea7c91f235fd32daeefdf875a0f938db05fb315"),
filepath.Join(p, "blobs", "sha256-bc80b03733773e0728011b2f4adf34c458b400e1aad48cb28d61170f3a2ad2d6"), filepath.Join(p, "blobs", "sha256-f6e7e4b28e0b1d0c635f2d465bd248c5387c3e75b61a48c4374192b26d832a56"),
}) })
w = createRequest(t, s.CreateHandler, api.CreateRequest{ w = createRequest(t, s.CreateHandler, api.CreateRequest{
@@ -227,8 +233,8 @@ func TestCreateRemovesLayers(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-8f2c2167d789c6b2302dff965160fa5029f6a24096d262c1cbb469f21a045382"), filepath.Join(p, "blobs", "sha256-136bf7c76bac2ec09d6617885507d37829e04b41acc47687d45e512b544e893a"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-fe7ac77b725cda2ccad03f88a880ecdfd7a33192d6cae08fce2c0ee1455991ed"), filepath.Join(p, "blobs", "sha256-fe7ac77b725cda2ccad03f88a880ecdfd7a33192d6cae08fce2c0ee1455991ed"),
}) })
} }
@@ -257,8 +263,8 @@ func TestCreateUnsetsSystem(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-8585df945d1069bc78b79bd10bb73ba07fbc29b0f5479a31a601c0d12731416e"), filepath.Join(p, "blobs", "sha256-0a666d113e8e0a3d27e9c7bd136a0bdfb6241037db50729d81568451ebfdbde8"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-f29e82a8284dbdf5910b1555580ff60b04238b8da9d5e51159ada67a4d0d5851"), filepath.Join(p, "blobs", "sha256-f29e82a8284dbdf5910b1555580ff60b04238b8da9d5e51159ada67a4d0d5851"),
}) })
@@ -278,8 +284,8 @@ func TestCreateUnsetsSystem(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-6bcdb8859d417753645538d7bbfbd7ca91a3f0c191aef5379c53c05e86b669dd"),
filepath.Join(p, "blobs", "sha256-ca239d7bd8ea90e4a5d2e6bf88f8d74a47b14336e73eb4e18bed4dd325018116"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
}) })
} }
@@ -312,8 +318,8 @@ func TestCreateMergeParameters(t *testing.T) {
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-1d0ad71299d48c2fb7ae2b98e683643e771f8a5b72be34942af90d97a91c1e37"), filepath.Join(p, "blobs", "sha256-1d0ad71299d48c2fb7ae2b98e683643e771f8a5b72be34942af90d97a91c1e37"),
filepath.Join(p, "blobs", "sha256-4a384beaf47a9cbe452dfa5ab70eea691790f3b35a832d12933a1996685bf2b6"), filepath.Join(p, "blobs", "sha256-6d6e36c1f90fc7deefc33a7300aa21ad4b67c506e33ecdeddfafa98147e60bbf"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
}) })
// in order to merge parameters, the second model must be created FROM the first // in order to merge parameters, the second model must be created FROM the first
@@ -354,9 +360,9 @@ func TestCreateMergeParameters(t *testing.T) {
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-1d0ad71299d48c2fb7ae2b98e683643e771f8a5b72be34942af90d97a91c1e37"), filepath.Join(p, "blobs", "sha256-1d0ad71299d48c2fb7ae2b98e683643e771f8a5b72be34942af90d97a91c1e37"),
filepath.Join(p, "blobs", "sha256-4a384beaf47a9cbe452dfa5ab70eea691790f3b35a832d12933a1996685bf2b6"), filepath.Join(p, "blobs", "sha256-6d6e36c1f90fc7deefc33a7300aa21ad4b67c506e33ecdeddfafa98147e60bbf"),
filepath.Join(p, "blobs", "sha256-4cd9d4ba6b734d9b4cbd1e5caa60374c00722e993fce5e1e2d15a33698f71187"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-bbdce269dabe013033632238b4b2d1e02fac2f97787c5e895f4da84e09cccd5d"),
filepath.Join(p, "blobs", "sha256-e29a7b3c47287a2489c895d21fe413c20f859a85d20e749492f52a838e36e1ba"), filepath.Join(p, "blobs", "sha256-e29a7b3c47287a2489c895d21fe413c20f859a85d20e749492f52a838e36e1ba"),
}) })
@@ -398,9 +404,9 @@ func TestCreateMergeParameters(t *testing.T) {
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-12f58bb75cb3042d69a7e013ab87fb3c3c7088f50ddc62f0c77bd332f0d44d35"), filepath.Join(p, "blobs", "sha256-12f58bb75cb3042d69a7e013ab87fb3c3c7088f50ddc62f0c77bd332f0d44d35"),
filepath.Join(p, "blobs", "sha256-1d0ad71299d48c2fb7ae2b98e683643e771f8a5b72be34942af90d97a91c1e37"), filepath.Join(p, "blobs", "sha256-1d0ad71299d48c2fb7ae2b98e683643e771f8a5b72be34942af90d97a91c1e37"),
filepath.Join(p, "blobs", "sha256-257aa726584f24970a4f240765e75a7169bfbe7f4966c1f04513d6b6c860583a"), filepath.Join(p, "blobs", "sha256-6d6e36c1f90fc7deefc33a7300aa21ad4b67c506e33ecdeddfafa98147e60bbf"),
filepath.Join(p, "blobs", "sha256-4a384beaf47a9cbe452dfa5ab70eea691790f3b35a832d12933a1996685bf2b6"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-9443591d14be23c1e33d101934d76ad03bdb0715fe0879e8b0d1819e7bb063dd"),
}) })
actual, err = os.ReadFile(filepath.Join(p, "blobs", "sha256-12f58bb75cb3042d69a7e013ab87fb3c3c7088f50ddc62f0c77bd332f0d44d35")) actual, err = os.ReadFile(filepath.Join(p, "blobs", "sha256-12f58bb75cb3042d69a7e013ab87fb3c3c7088f50ddc62f0c77bd332f0d44d35"))
@@ -456,8 +462,8 @@ func TestCreateReplacesMessages(t *testing.T) {
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-298baeaf6928a60cf666d88d64a1ba606feb43a2865687c39e40652e407bffc4"), filepath.Join(p, "blobs", "sha256-298baeaf6928a60cf666d88d64a1ba606feb43a2865687c39e40652e407bffc4"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-e0e27d47045063ccb167ae852c51d49a98eab33fabaee4633fdddf97213e40b5"), filepath.Join(p, "blobs", "sha256-c84aee28f2af350596f674de51d2a802ea782653ef2930a21d48bd43d5cd5317"),
}) })
w = createRequest(t, s.CreateHandler, api.CreateRequest{ w = createRequest(t, s.CreateHandler, api.CreateRequest{
@@ -491,11 +497,11 @@ func TestCreateReplacesMessages(t *testing.T) {
// Old layers will not have been pruned // Old layers will not have been pruned
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-09cfac3e6a637e25cb41aa85c24c110dc17ba89634de7df141b564dd2da4168b"),
filepath.Join(p, "blobs", "sha256-298baeaf6928a60cf666d88d64a1ba606feb43a2865687c39e40652e407bffc4"), filepath.Join(p, "blobs", "sha256-298baeaf6928a60cf666d88d64a1ba606feb43a2865687c39e40652e407bffc4"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-a60ecc9da299ec7ede453f99236e5577fd125e143689b646d9f0ddc9971bf4db"), filepath.Join(p, "blobs", "sha256-a60ecc9da299ec7ede453f99236e5577fd125e143689b646d9f0ddc9971bf4db"),
filepath.Join(p, "blobs", "sha256-e0e27d47045063ccb167ae852c51d49a98eab33fabaee4633fdddf97213e40b5"), filepath.Join(p, "blobs", "sha256-c84aee28f2af350596f674de51d2a802ea782653ef2930a21d48bd43d5cd5317"),
filepath.Join(p, "blobs", "sha256-f4e2c3690efef1b4b63ba1e1b2744ffeb6a7438a0110b86596069f6d9999c80b"),
}) })
type message struct { type message struct {
@@ -550,9 +556,9 @@ func TestCreateTemplateSystem(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-2b5e330885117c82f3fd75169ea323e141070a2947c11ddb9f79ee0b01c589c1"), filepath.Join(p, "blobs", "sha256-0a04d979734167da3b80811a1874d734697f366a689f3912589b99d2e86e7ad1"),
filepath.Join(p, "blobs", "sha256-4c5f51faac758fecaff8db42f0b7382891a4d0c0bb885f7b86be88c814a7cc86"), filepath.Join(p, "blobs", "sha256-4c5f51faac758fecaff8db42f0b7382891a4d0c0bb885f7b86be88c814a7cc86"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-fe7ac77b725cda2ccad03f88a880ecdfd7a33192d6cae08fce2c0ee1455991ed"), filepath.Join(p, "blobs", "sha256-fe7ac77b725cda2ccad03f88a880ecdfd7a33192d6cae08fce2c0ee1455991ed"),
}) })
@@ -714,8 +720,8 @@ func TestCreateLicenses(t *testing.T) {
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-2af71558e438db0b73a20beab92dc278a94e1bbe974c00c1a33e3ab62d53a608"), filepath.Join(p, "blobs", "sha256-2af71558e438db0b73a20beab92dc278a94e1bbe974c00c1a33e3ab62d53a608"),
filepath.Join(p, "blobs", "sha256-79a39c37536ddee29cbadd5d5e2dcba8ed7f03e431f626ff38432c1c866bb7e2"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-a762f214df0d96c9a7b82f96da98d99ceb2776c88e3ea7ffa09d1e5835516ec6"),
filepath.Join(p, "blobs", "sha256-e5dcffe836b6ec8a58e492419b550e65fb8cbdc308503979e5dacb33ac7ea3b7"), filepath.Join(p, "blobs", "sha256-e5dcffe836b6ec8a58e492419b550e65fb8cbdc308503979e5dacb33ac7ea3b7"),
}) })
@@ -761,9 +767,9 @@ func TestCreateDetectTemplate(t *testing.T) {
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-0d79f567714c62c048378f2107fb332dabee0135d080c302d884317da9433cc5"), filepath.Join(p, "blobs", "sha256-0d79f567714c62c048378f2107fb332dabee0135d080c302d884317da9433cc5"),
filepath.Join(p, "blobs", "sha256-3322a0c650c758b7386ff55629d27d07c07b6c3d3515e259dc3e5598c41e9f4e"),
filepath.Join(p, "blobs", "sha256-35360843d0c84fb1506952a131bbef13cd2bb4a541251f22535170c05b56e672"), filepath.Join(p, "blobs", "sha256-35360843d0c84fb1506952a131bbef13cd2bb4a541251f22535170c05b56e672"),
filepath.Join(p, "blobs", "sha256-553c4a3f747b3d22a4946875f1cc8ed011c2930d83f864a0c7265f9ec0a20413"), filepath.Join(p, "blobs", "sha256-a56c12acca8068cb6c335e237da6643e8a802a92959a63ad5bd17828e3b5e9b0"),
filepath.Join(p, "blobs", "sha256-de3959f841e9ef6b4b6255fa41cb9e0a45da89c3066aa72bdd07a4747f848990"),
}) })
}) })
@@ -780,8 +786,8 @@ func TestCreateDetectTemplate(t *testing.T) {
} }
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-6bcdb8859d417753645538d7bbfbd7ca91a3f0c191aef5379c53c05e86b669dd"),
filepath.Join(p, "blobs", "sha256-ca239d7bd8ea90e4a5d2e6bf88f8d74a47b14336e73eb4e18bed4dd325018116"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
}) })
}) })
} }

10
server/routes_delete_test.go
View File

@@ -47,9 +47,9 @@ func TestDelete(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-8f2c2167d789c6b2302dff965160fa5029f6a24096d262c1cbb469f21a045382"), filepath.Join(p, "blobs", "sha256-136bf7c76bac2ec09d6617885507d37829e04b41acc47687d45e512b544e893a"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-6bcdb8859d417753645538d7bbfbd7ca91a3f0c191aef5379c53c05e86b669dd"),
filepath.Join(p, "blobs", "sha256-ca239d7bd8ea90e4a5d2e6bf88f8d74a47b14336e73eb4e18bed4dd325018116"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-fe7ac77b725cda2ccad03f88a880ecdfd7a33192d6cae08fce2c0ee1455991ed"), filepath.Join(p, "blobs", "sha256-fe7ac77b725cda2ccad03f88a880ecdfd7a33192d6cae08fce2c0ee1455991ed"),
}) })
@@ -64,8 +64,8 @@ func TestDelete(t *testing.T) {
}) })
checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{ checkFileExists(t, filepath.Join(p, "blobs", "*"), []string{
filepath.Join(p, "blobs", "sha256-8f2c2167d789c6b2302dff965160fa5029f6a24096d262c1cbb469f21a045382"), filepath.Join(p, "blobs", "sha256-136bf7c76bac2ec09d6617885507d37829e04b41acc47687d45e512b544e893a"),
filepath.Join(p, "blobs", "sha256-a4e5e156ddec27e286f75328784d7106b60a4eb1d246e950a001a3f944fbda99"), filepath.Join(p, "blobs", "sha256-89a2116c3a82d6a97f59f748d86ed4417214353fd178ee54df418fde32495fad"),
filepath.Join(p, "blobs", "sha256-fe7ac77b725cda2ccad03f88a880ecdfd7a33192d6cae08fce2c0ee1455991ed"), filepath.Join(p, "blobs", "sha256-fe7ac77b725cda2ccad03f88a880ecdfd7a33192d6cae08fce2c0ee1455991ed"),
}) })