Files
ollama37/integration/tools_test.go
Shang Chieh Tseng ef14fb5b26 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- Removed outdated CI/CD workflows (tesla-k80-*.yml)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>
2025-11-05 14:03:05 +08:00

133 lines
3.1 KiB
Go

//go:build integration
package integration
import (
"context"
"fmt"
"testing"
"time"
"github.com/ollama/ollama/api"
)
func TestAPIToolCalling(t *testing.T) {
initialTimeout := 60 * time.Second
streamTimeout := 60 * time.Second
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Minute)
defer cancel()
client, _, cleanup := InitServerConnection(ctx, t)
defer cleanup()
minVRAM := map[string]uint64{
"qwen3-vl": 16,
"gpt-oss:20b": 16,
"gpt-oss:120b": 70,
"qwen3": 6,
"llama3.1": 8,
"llama3.2": 4,
"mistral": 6,
"qwen2.5": 6,
"qwen2": 6,
"mistral-nemo": 9,
"mistral-small": 16,
"mixtral:8x22b": 80,
"qwq": 20,
"granite3.3": 7,
}
for _, model := range libraryToolsModels {
t.Run(model, func(t *testing.T) {
if v, ok := minVRAM[model]; ok {
skipUnderMinVRAM(t, v)
}
if err := PullIfMissing(ctx, client, model); err != nil {
t.Fatalf("pull failed %s", err)
}
tools := []api.Tool{
{
Type: "function",
Function: api.ToolFunction{
Name: "get_weather",
Description: "Get the current weather in a given location",
Parameters: api.ToolFunctionParameters{
Type: "object",
Required: []string{"location"},
Properties: map[string]api.ToolProperty{
"location": {
Type: api.PropertyType{"string"},
Description: "The city and state, e.g. San Francisco, CA",
},
},
},
},
},
}
req := api.ChatRequest{
Model: model,
Messages: []api.Message{
{
Role: "user",
Content: "Call get_weather with location set to San Francisco.",
},
},
Tools: tools,
Options: map[string]any{
"temperature": 0,
},
}
stallTimer := time.NewTimer(initialTimeout)
var gotToolCall bool
var lastToolCall api.ToolCall
fn := func(response api.ChatResponse) error {
if len(response.Message.ToolCalls) > 0 {
gotToolCall = true
lastToolCall = response.Message.ToolCalls[len(response.Message.ToolCalls)-1]
}
if !stallTimer.Reset(streamTimeout) {
return fmt.Errorf("stall was detected while streaming response, aborting")
}
return nil
}
stream := true
req.Stream = &stream
done := make(chan int)
var genErr error
go func() {
genErr = client.Chat(ctx, &req, fn)
done <- 0
}()
select {
case <-stallTimer.C:
t.Errorf("tool-calling chat never started. Timed out after: %s", initialTimeout.String())
case <-done:
if genErr != nil {
t.Fatalf("chat failed: %v", genErr)
}
if !gotToolCall {
t.Fatalf("expected at least one tool call, got none")
}
if lastToolCall.Function.Name != "get_weather" {
t.Errorf("unexpected tool called: got %q want %q", lastToolCall.Function.Name, "get_weather")
}
if _, ok := lastToolCall.Function.Arguments["location"]; !ok {
t.Errorf("expected tool arguments to include 'location', got: %s", lastToolCall.Function.Arguments.String())
}
case <-ctx.Done():
t.Error("outer test context done while waiting for tool-calling chat")
}
})
}
}