mirror of
https://github.com/ollama/ollama.git
synced 2025-03-21 15:23:01 +01:00
ed443a0393
This provides integration with the new Ollama engine (5824541 next ollama runner (#7913)) and the rest of the Ollama infrastructure such as the runner and Ollama server. In addition, it also builds out the KV cache infrastructure to support requirements of how Ollama runs models such as: - Parallel processing - Memory management for defragmentation and shifting - Multi-modal modals Both old and new engines continue to be supported. By default, only the old engine is used. To enable the new engine: Start the server with the OLLAMA_NEW_ENGINE environment variable set: OLLAMA_NEW_ENGINE=1 ./ollama serve Start a model that is supported by the Ollama engine. This one is Llama 3.1 8b Q4_K_M: ./ollama run jessegross/llama3.1
97 lines
1.9 KiB
Go
97 lines
1.9 KiB
Go
package common
|
|
|
|
import (
|
|
"strings"
|
|
)
|
|
|
|
func FindStop(sequence string, stops []string) (bool, string) {
|
|
for _, stop := range stops {
|
|
if strings.Contains(sequence, stop) {
|
|
return true, stop
|
|
}
|
|
}
|
|
|
|
return false, ""
|
|
}
|
|
|
|
func ContainsStopSuffix(sequence string, stops []string) bool {
|
|
for _, stop := range stops {
|
|
for i := 1; i <= len(stop); i++ {
|
|
if strings.HasSuffix(sequence, stop[:i]) {
|
|
return true
|
|
}
|
|
}
|
|
}
|
|
|
|
return false
|
|
}
|
|
|
|
// truncateStop removes the provided stop string from pieces,
|
|
// returning the partial pieces with stop removed, including truncating
|
|
// the last piece if required (and signalling if this was the case)
|
|
func TruncateStop(pieces []string, stop string) ([]string, bool) {
|
|
joined := strings.Join(pieces, "")
|
|
|
|
index := strings.Index(joined, stop)
|
|
if index == -1 {
|
|
return pieces, false
|
|
}
|
|
|
|
joined = joined[:index]
|
|
|
|
// Split truncated string back into pieces of original lengths
|
|
lengths := make([]int, len(pieces))
|
|
for i, piece := range pieces {
|
|
lengths[i] = len(piece)
|
|
}
|
|
|
|
var result []string
|
|
tokenTruncated := false
|
|
start := 0
|
|
for _, length := range lengths {
|
|
if start >= len(joined) {
|
|
break
|
|
}
|
|
|
|
end := start + length
|
|
if end > len(joined) {
|
|
end = len(joined)
|
|
tokenTruncated = true
|
|
}
|
|
result = append(result, joined[start:end])
|
|
start = end
|
|
}
|
|
|
|
return result, tokenTruncated
|
|
}
|
|
|
|
func IncompleteUnicode(token string) bool {
|
|
incomplete := false
|
|
|
|
// check if there is incomplete UTF-8 character at the end
|
|
for i := 1; i < 5 && i <= len(token); i++ {
|
|
c := token[len(token)-i]
|
|
|
|
if (c & 0xc0) == 0x80 {
|
|
// continuation byte: 10xxxxxx
|
|
continue
|
|
}
|
|
|
|
if (c & 0xe0) == 0xc0 {
|
|
// 2-byte character: 110xxxxx ...
|
|
incomplete = i < 2
|
|
} else if (c & 0xf0) == 0xe0 {
|
|
// 3-byte character: 1110xxxx ...
|
|
incomplete = i < 3
|
|
} else if (c & 0xf8) == 0xf0 {
|
|
// 4-byte character: 11110xxx ...
|
|
incomplete = i < 4
|
|
}
|
|
|
|
// else 1-byte character or invalid byte
|
|
break
|
|
}
|
|
|
|
return incomplete
|
|
}
|