mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-11 00:07:07 +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with comprehensive build instructions - Documented toolchain constraints and CPU architecture trade-offs - Removed outdated CI/CD workflows (tesla-k80-*.yml) - Cleaned up temporary development artifacts ## Rationale This fork maintains Tesla K80 GPU support (compute 3.7) which was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>
66 lines
1.3 KiB
Go
66 lines
1.3 KiB
Go
package rope
|
|
|
|
import "github.com/ollama/ollama/ml"
|
|
|
|
// Options contains optional parameters for RoPE function
|
|
type Options struct {
|
|
Type int
|
|
Factors ml.Tensor
|
|
|
|
// YaRN options
|
|
YaRN struct {
|
|
OriginalContextLength int
|
|
ExtrapolationFactor,
|
|
AttentionFactor,
|
|
BetaFast,
|
|
BetaSlow float32
|
|
}
|
|
|
|
// MRoPE options
|
|
MRoPE struct {
|
|
Sections []int
|
|
}
|
|
}
|
|
|
|
// WithTypeNeoX sets RoPE type to NeoX
|
|
func WithTypeNeoX() func(*Options) {
|
|
return func(opts *Options) {
|
|
opts.Type = 2
|
|
}
|
|
}
|
|
|
|
// WithFactors sets custom rope factors
|
|
func WithFactors(factors ml.Tensor) func(*Options) {
|
|
return func(opts *Options) {
|
|
if factors != nil {
|
|
opts.Factors = factors
|
|
}
|
|
}
|
|
}
|
|
|
|
// WithOriginalContextLength sets a custom context length
|
|
func WithOriginalContextLength(n int) func(*Options) {
|
|
return func(opts *Options) {
|
|
opts.YaRN.OriginalContextLength = n
|
|
}
|
|
}
|
|
|
|
func WithExtrapolationFactor(extrapolationFactor float32) func(*Options) {
|
|
return func(opts *Options) {
|
|
opts.YaRN.ExtrapolationFactor = extrapolationFactor
|
|
}
|
|
}
|
|
|
|
func WithAttentionFactor(attentionFactor float32) func(*Options) {
|
|
return func(opts *Options) {
|
|
opts.YaRN.AttentionFactor = attentionFactor
|
|
}
|
|
}
|
|
|
|
func WithMRoPESections(sections []int) func(*Options) {
|
|
return func(opts *Options) {
|
|
opts.Type |= 1 << 3
|
|
opts.MRoPE.Sections = sections
|
|
}
|
|
}
|