mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-11 08:17:03 +00:00
ef14fb5b26
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with comprehensive build instructions - Documented toolchain constraints and CPU architecture trade-offs - Removed outdated CI/CD workflows (tesla-k80-*.yml) - Cleaned up temporary development artifacts ## Rationale This fork maintains Tesla K80 GPU support (compute 3.7) which was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>
65 lines
1.4 KiB
Go
65 lines
1.4 KiB
Go
package pooling_test
|
|
|
|
import (
|
|
"bytes"
|
|
"os"
|
|
"testing"
|
|
|
|
"github.com/google/go-cmp/cmp"
|
|
fsggml "github.com/ollama/ollama/fs/ggml"
|
|
"github.com/ollama/ollama/ml"
|
|
"github.com/ollama/ollama/ml/backend/ggml"
|
|
"github.com/ollama/ollama/ml/nn/pooling"
|
|
)
|
|
|
|
func setup(tb testing.TB, n int) ml.Backend {
|
|
tb.Helper()
|
|
|
|
f, err := os.CreateTemp(tb.TempDir(), "*.bin")
|
|
if err != nil {
|
|
tb.Fatal(err)
|
|
}
|
|
defer f.Close()
|
|
|
|
if err := fsggml.WriteGGUF(f, fsggml.KV{
|
|
"general.architecture": "test",
|
|
"test.block_count": uint32(1),
|
|
}, []*fsggml.Tensor{
|
|
{Name: "blk.0.weight", Shape: []uint64{1}, WriterTo: bytes.NewBuffer(make([]byte, 4))},
|
|
}); err != nil {
|
|
tb.Fatal(err)
|
|
}
|
|
|
|
b, err := ggml.New(f.Name(), ml.BackendParams{AllocMemory: true})
|
|
if err != nil {
|
|
tb.Fatal(err)
|
|
}
|
|
|
|
return b
|
|
}
|
|
|
|
func TestForward(t *testing.T) {
|
|
cases := map[pooling.Type][]float32{
|
|
pooling.TypeMean: {4, 5, 6, 7, 8, 9, 10, 11},
|
|
pooling.TypeCLS: {0, 1, 2, 3, 4, 5, 6, 7},
|
|
pooling.TypeLast: {8, 9, 10, 11, 12, 13, 14, 15},
|
|
}
|
|
for typ, want := range cases {
|
|
t.Run(typ.String(), func(t *testing.T) {
|
|
b := setup(t, 99)
|
|
defer b.Close()
|
|
|
|
ctx := b.NewContext()
|
|
defer ctx.Close()
|
|
|
|
tt := ctx.Input().Arange(0, 16, 1, ml.DTypeF32).Reshape(ctx, 8, 2)
|
|
tt = typ.Forward(ctx, tt)
|
|
|
|
ctx.Forward(tt).Compute(tt)
|
|
if diff := cmp.Diff(want, tt.Floats()); diff != "" {
|
|
t.Error(diff)
|
|
}
|
|
})
|
|
}
|
|
}
|