matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-09 23:37:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

model: handle multiple eos tokens (#10577)

Browse Source 
* get eos_token_id from generation_config.json

* refactor

* include both ids and strings in trace

* comments

* remove special case for gemma3 special vocab (#10743)
This commit is contained in:
Michael Yang
2025-05-16 13:40:23 -07:00
committed by GitHub
parent 27da2cddc5
commit 333e360422
18 changed files with 282 additions and 182 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
llama/llama.go
View File

@@ -602,7 +602,7 @@ type Grammar struct {
mu sync.Mutex
}
func NewGrammar(grammar string, vocabIds []uint32, vocabValues []string, eogTokens []uint32) *Grammar {
func NewGrammar(grammar string, vocabIds []uint32, vocabValues []string, eogTokens []int32) *Grammar {
cGrammar := C.CString(grammar)
defer C.free(unsafe.Pointer(cGrammar))
@@ -622,7 +622,7 @@ func NewGrammar(grammar string, vocabIds []uint32, vocabValues []string, eogToke
cEogTokens[i] = C.uint32_t(token)
}
g := C.grammar_init(cGrammar, (*C.uint32_t)(unsafe.Pointer(&cTokens[0])), C.size_t(len(cTokens)), (**C.char)(unsafe.Pointer(&cPieces[0])), (*C.uint32_t)(unsafe.Pointer(&cEogTokens[0])), C.size_t(len(cEogTokens)))
g := C.grammar_init(cGrammar, unsafe.SliceData(cTokens), C.size_t(len(cTokens)), unsafe.SliceData(cPieces), unsafe.SliceData(cEogTokens), C.size_t(len(cEogTokens)))
if g == nil {
return nil
}