starred/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-04-22 00:36:11 +02:00
Code Issues Packages Projects Releases Wiki Activity

merge ggml file decoding

Browse Source
This commit is contained in:
Michael Yang
2024-12-03 11:17:45 -08:00
parent 2c5fb24855
commit b7943d941d
30 changed files with 429 additions and 1733 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
convert/convert_phi3.go
View File

@@ -8,7 +8,7 @@ import (
"strings"
"sync"
"github.com/ollama/ollama/llm"
"github.com/ollama/ollama/fs/ggml"
)
type phi3Model struct {
@@ -37,7 +37,7 @@ type phi3Model struct {
var _ ModelConverter = (*phi3Model)(nil)
func (p *phi3Model) KV(t *Tokenizer) llm.KV {
func (p *phi3Model) KV(t *Tokenizer) ggml.KV {
kv := p.ModelParameters.KV(t)
kv["general.architecture"] = "phi3"
kv["phi3.context_length"] = p.MaxPositionEmbeddings
@@ -68,19 +68,19 @@ func (p *phi3Model) KV(t *Tokenizer) llm.KV {
return kv
}
func (p *phi3Model) Tensors(ts []Tensor) []llm.Tensor {
func (p *phi3Model) Tensors(ts []Tensor) []ggml.Tensor {
var addRopeFactors sync.Once
out := make([]llm.Tensor, 0, len(ts)+2)
out := make([]ggml.Tensor, 0, len(ts)+2)
for _, t := range ts {
if strings.HasPrefix(t.Name(), "blk.0.") {
addRopeFactors.Do(func() {
out = append(out, llm.Tensor{
out = append(out, ggml.Tensor{
Name: "rope_factors_long.weight",
Kind: 0,
Shape: []uint64{uint64(len(p.RopeScaling.LongFactor))},
WriterTo: p.RopeScaling.LongFactor,
}, llm.Tensor{
}, ggml.Tensor{
Name: "rope_factors_short.weight",
Kind: 0,
Shape: []uint64{uint64(len(p.RopeScaling.ShortFactor))},
@@ -89,7 +89,7 @@ func (p *phi3Model) Tensors(ts []Tensor) []llm.Tensor {
})
}
out = append(out, llm.Tensor{
out = append(out, ggml.Tensor{
Name: t.Name(),
Kind: t.Kind(),
Shape: t.Shape(),