mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-09 23:37:06 +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with comprehensive build instructions - Documented toolchain constraints and CPU architecture trade-offs - Removed outdated CI/CD workflows (tesla-k80-*.yml) - Cleaned up temporary development artifacts ## Rationale This fork maintains Tesla K80 GPU support (compute 3.7) which was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>
52 lines
1.2 KiB
Go
52 lines
1.2 KiB
Go
package model
|
|
|
|
import (
|
|
"slices"
|
|
"testing"
|
|
|
|
"github.com/google/go-cmp/cmp"
|
|
)
|
|
|
|
func TestWordPiece(t *testing.T) {
|
|
wpm := NewWordPiece(
|
|
&Vocabulary{
|
|
Values: []string{"[UNK]", "[CLS]", "[SEP]", "▁hello", "▁world", "s", "▁!", "▁@", "▁#"},
|
|
AddBOS: true,
|
|
AddEOS: true,
|
|
BOS: []int32{1},
|
|
EOS: []int32{2},
|
|
})
|
|
|
|
ids, err := wpm.Encode("Hello world!", true)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
|
|
if diff := cmp.Diff([]int32{1, 3, 4, 6, 2}, ids); diff != "" {
|
|
t.Errorf("unexpected ids (-want +got):\n%s", diff)
|
|
}
|
|
|
|
words, err := wpm.Decode(ids)
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
|
|
if diff := cmp.Diff("[CLS] hello world! [SEP]", words); diff != "" {
|
|
t.Errorf("unexpected words (-want +got):\n%s", diff)
|
|
}
|
|
}
|
|
|
|
func TestWordPieceWords(t *testing.T) {
|
|
var wpm WordPiece
|
|
|
|
basic := slices.Collect(wpm.words("Hey friend! How are you?!?"))
|
|
if diff := cmp.Diff([]string{"Hey", "friend", "!", "How", "are", "you", "?", "!", "?"}, basic); diff != "" {
|
|
t.Errorf("unexpected words (-want +got):\n%s", diff)
|
|
}
|
|
|
|
chinese := slices.Collect(wpm.words("野口里佳 Noguchi Rika"))
|
|
if diff := cmp.Diff([]string{"野", "口", "里", "佳", "Noguchi", "Rika"}, chinese); diff != "" {
|
|
t.Errorf("unexpected words (-want +got):\n%s", diff)
|
|
}
|
|
}
|