matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-10 15:57:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

Move quantization to new backend (#10363)

Browse Source 
* Move quantization logic to GGML via new backend

This moves the model aware logic to Go code and calls GGMLs quantization code for model creation.

* Remove "add model quantizations"

This is no longer needed now that quantization is implemented in Go+GGML code directly.
This commit is contained in:
Daniel Hiltgen
2025-05-06 11:20:48 -07:00
committed by GitHub
parent 95e744beeb
commit 424810450f
39 changed files with 1854 additions and 440 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
convert/convert_llama.go
View File

@@ -126,11 +126,11 @@ func (p *llamaModel) KV(t *Tokenizer) ggml.KV {
return kv
}
func (p *llamaModel) Tensors(ts []Tensor) []ggml.Tensor {
var out []ggml.Tensor
func (p *llamaModel) Tensors(ts []Tensor) []*ggml.Tensor {
var out []*ggml.Tensor
if p.RopeScaling.factors != nil {
out = append(out, ggml.Tensor{
out = append(out, &ggml.Tensor{
Name: "rope_freqs.weight",
Kind: 0,
Shape: []uint64{uint64(len(p.RopeScaling.factors))},
@@ -145,7 +145,7 @@ func (p *llamaModel) Tensors(ts []Tensor) []ggml.Tensor {
}
}
out = append(out, ggml.Tensor{
out = append(out, &ggml.Tensor{
Name: t.Name(),
Kind: t.Kind(),
Shape: t.Shape(),