mirror of
https://github.com/ollama/ollama.git
synced 2026-03-28 03:08:44 +07:00
9667c2282f
TeaCache: - Timestep embedding similarity caching for diffusion models - Polynomial rescaling with configurable thresholds - Reduces transformer forward passes by ~30-50% FP8 quantization: - Support for FP8 quantized models (8-bit weights with scales) - QuantizedMatmul on Metal, Dequantize on CUDA - Client-side quantization via ollama create --quantize fp8 Other bug fixes: - Fix `/api/show` API for image generation models - Server properly returns model info (architecture, parameters, quantization) - Memory allocation optimizations - CLI improvements for image generation
19 lines
496 B
Go
19 lines
496 B
Go
//go:build !mlx
|
|
|
|
package client
|
|
|
|
import (
|
|
"fmt"
|
|
"io"
|
|
)
|
|
|
|
// quantizeTensor is not available without MLX
|
|
func quantizeTensor(r io.Reader, name, dtype string, shape []int32) (qweightData, scalesData, qbiasData []byte, qweightShape, scalesShape, qbiasShape []int32, err error) {
|
|
return nil, nil, nil, nil, nil, nil, fmt.Errorf("quantization requires MLX support (build with mlx tag)")
|
|
}
|
|
|
|
// QuantizeSupported returns false when MLX is not available
|
|
func QuantizeSupported() bool {
|
|
return false
|
|
}
|