highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-11-11 16:27:02 +01:00
Code Issues Packages Projects Releases Wiki Activity

model: implement bert in ollama engine (#9080)

Browse Source 
* fix truncate

* s/SentencePieceModel/SentencePiece/

* bert

* wordpiece

* refactor pooling

* more tokenizers

* normalize embeddings
This commit is contained in:
Michael Yang
2025-09-15 15:35:59 -07:00
committed by GitHub
parent 6f7117145f
commit 3f6642f6fc
16 changed files with 490 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
server/routes.go
View File

@@ -488,7 +488,6 @@ func (s *Server) EmbedHandler(c *gin.Context) {
}
truncate := true
if req.Truncate != nil && !*req.Truncate {
truncate = false
}
@@ -555,7 +554,16 @@ func (s *Server) EmbedHandler(c *gin.Context) {
return
}
if bos := kvData.Uint("tokenizer.ggml.bos_token_id"); tokens[0] != int(bos) && kvData.Bool("add_bos_token", true) {
ctxLen--
}
if eos := kvData.Uint("tokenizer.ggml.eos_token_id"); tokens[len(tokens)-1] != int(eos) && kvData.Bool("add_eos_token", true) {
ctxLen--
}
tokens = tokens[:ctxLen]
s, err = r.Detokenize(c.Request.Context(), tokens)
if err != nil {
c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})