highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-03-28 02:33:14 +01:00
Code Issues Packages Projects Releases Wiki Activity

add command-r graph estimate

Browse Source
This commit is contained in:
Michael Yang 2024-04-04 13:24:20 -07:00
parent 483b81a863
commit 01f77ae25d
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

@ -327,7 +327,7 @@ func (llm GGML) GraphSize(context, batch int) (int64, bool) {
}
switch llm.KV().Architecture() {
case "gemma":
case "gemma", "command-r":
return 4 * int64(batch) * int64(embeddingLength+uint64(vocabLength)), true
case "phi2":
return max(