Sync with upstream ollama/ollama and restore Tesla K80 (compute 3.7) support

This commit represents a complete rework after pulling the latest changes from
official ollama/ollama repository and re-applying Tesla K80 compatibility patches.

## Key Changes

### CUDA Compute Capability 3.7 Support (Tesla K80)
- Added sm_37 (compute 3.7) to CMAKE_CUDA_ARCHITECTURES in CMakeLists.txt
- Updated CMakePresets.json to include compute 3.7 in "CUDA 11" preset
- Using 37-virtual (PTX with JIT compilation) for maximum compatibility

### Legacy Toolchain Compatibility
- **NVIDIA Driver**: 470.256.02 (last version supporting Kepler/K80)
- **CUDA Version**: 11.4.4 (last CUDA 11.x supporting compute 3.7)
- **GCC Version**: 10.5.0 (required by CUDA 11.4 host_config.h)

### CPU Architecture Trade-offs
Due to GCC 10.5 limitation, sacrificed newer CPU optimizations:
- Alderlake CPU variant enabled WITHOUT AVX_VNNI (requires GCC 11+)
- Still supports: SSE4.2, AVX, F16C, AVX2, BMI2, FMA
- Performance impact: ~3-7% on newer CPUs (acceptable for K80 compatibility)

### Build System Updates
- Modified ml/backend/ggml/ggml/src/ggml-cuda/CMakeLists.txt for compute 3.7
- Added -Wno-deprecated-gpu-targets flag to suppress warnings
- Updated ml/backend/ggml/ggml/src/CMakeLists.txt for Alderlake without AVX_VNNI

### Upstream Sync
Merged latest llama.cpp changes including:
- Enhanced KV cache management with ISWA and hybrid memory support
- Improved multi-modal support (mtmd framework)
- New model architectures (Gemma3, Llama4, Qwen3, etc.)
- GPU backend improvements for CUDA, Metal, and ROCm
- Updated quantization support and GGUF format handling

### Documentation
- Updated CLAUDE.md with comprehensive build instructions
- Documented toolchain constraints and CPU architecture trade-offs
- Removed outdated CI/CD workflows (tesla-k80-*.yml)
- Cleaned up temporary development artifacts

## Rationale

This fork maintains Tesla K80 GPU support (compute 3.7) which was dropped in
official Ollama due to legacy driver/CUDA requirements. The toolchain constraint
creates a deadlock:
- K80 → Driver 470 → CUDA 11.4 → GCC 10 → No AVX_VNNI

We accept the loss of cutting-edge CPU optimizations to enable running modern
LLMs on legacy but still capable Tesla K80 hardware (12GB VRAM per GPU).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Shang Chieh Tseng
2025-11-05 14:03:05 +08:00
parent fabe2c5cb7
commit ef14fb5b26
817 changed files with 241634 additions and 70888 deletions

View File

@@ -17,6 +17,11 @@ func TestKeepAliveParsingFromJSON(t *testing.T) {
req string
exp *Duration
}{
{
name: "Unset",
req: `{ }`,
exp: nil,
},
{
name: "Positive Integer",
req: `{ "keep_alive": 42 }`,
@@ -25,7 +30,7 @@ func TestKeepAliveParsingFromJSON(t *testing.T) {
{
name: "Positive Float",
req: `{ "keep_alive": 42.5 }`,
exp: &Duration{42 * time.Second},
exp: &Duration{42500 * time.Millisecond},
},
{
name: "Positive Integer String",
@@ -293,6 +298,30 @@ func TestToolFunction_UnmarshalJSON(t *testing.T) {
}
}
func TestToolCallFunction_IndexAlwaysMarshals(t *testing.T) {
fn := ToolCallFunction{
Name: "echo",
Arguments: ToolCallFunctionArguments{"message": "hi"},
}
data, err := json.Marshal(fn)
require.NoError(t, err)
raw := map[string]any{}
require.NoError(t, json.Unmarshal(data, &raw))
require.Contains(t, raw, "index")
assert.Equal(t, float64(0), raw["index"])
fn.Index = 3
data, err = json.Marshal(fn)
require.NoError(t, err)
raw = map[string]any{}
require.NoError(t, json.Unmarshal(data, &raw))
require.Contains(t, raw, "index")
assert.Equal(t, float64(3), raw["index"])
}
func TestPropertyType_UnmarshalJSON(t *testing.T) {
tests := []struct {
name string
@@ -436,3 +465,50 @@ func TestThinking_UnmarshalJSON(t *testing.T) {
})
}
}
func TestToolFunctionParameters_String(t *testing.T) {
tests := []struct {
name string
params ToolFunctionParameters
expected string
}{
{
name: "simple object with string property",
params: ToolFunctionParameters{
Type: "object",
Required: []string{"name"},
Properties: map[string]ToolProperty{
"name": {
Type: PropertyType{"string"},
Description: "The name of the person",
},
},
},
expected: `{"type":"object","required":["name"],"properties":{"name":{"type":"string","description":"The name of the person"}}}`,
},
{
name: "marshal failure returns empty string",
params: ToolFunctionParameters{
Type: "object",
Defs: func() any {
// Create a cycle that will cause json.Marshal to fail
type selfRef struct {
Self *selfRef
}
s := &selfRef{}
s.Self = s
return s
}(),
Properties: map[string]ToolProperty{},
},
expected: "",
},
}
for _, test := range tests {
t.Run(test.name, func(t *testing.T) {
result := test.params.String()
assert.Equal(t, test.expected, result)
})
}
}