baalajimaestro/ollama
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

fix q5_0, q5_1

Browse source
This commit is contained in:
Michael Yang 2024-05-24 16:01:37 -07:00
parent 4cc3be3035
commit 8f440d579a
1 changed files with 1 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
llm/ggml.go
View file

@ -127,7 +127,7 @@ func (t Tensor) blockSize() uint64 {
switch t.Kind {
case 0, 1, 24, 25, 26, 27, 28, 31: // F32, F16, I8, I16, I32, I64, F64, BF16
return 1
case 2, 3, 8, 9, 20: // Q4_0, Q4_1, Q8_0, Q8_1, IQ4_NL
case 2, 3, 4, 5, 6, 7, 8, 9, 20: // Q4_0, Q4_1, Q5_0, Q5_1, Q8_0, Q8_1, IQ4_NL
return 32
default: // All others
return 256