haraldwolff/llama.cpp
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases Wiki Activity

swift : fix token_to_piece implementation (#4278)

Browse source 
* Fix token_to_piece implementation in Swift

* Fix errors
This commit is contained in:
Miwa / Ensan 2023-12-02 03:19:45 +09:00 committed by GitHub
parent 511f52c334
commit b220222a64
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 18 additions and 14 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
examples/batched.swift/Sources/main.swift
View file

@ -230,18 +230,15 @@ private func token_to_piece(token: llama_token, buffer: inout [CChar]) -> String
var result = [CChar](repeating: 0, count: 8)
let nTokens = llama_token_to_piece(model, token, &result, Int32(result.count))
if nTokens < 0 {
if result.count >= -Int(nTokens) {
result.removeLast(-Int(nTokens))
} else {
result.removeAll()
}
let actualTokensCount = -Int(nTokens)
result = .init(repeating: 0, count: actualTokensCount)
let check = llama_token_to_piece(
model,
token,
&result,
Int32(result.count)
)
assert(check == nTokens)
assert(check == actualTokensCount)
} else {
result.removeLast(result.count - Int(nTokens))
}
@ -259,5 +256,4 @@ private func token_to_piece(token: llama_token, buffer: inout [CChar]) -> String
buffer = []
return bufferString
}
return nil
}

22
examples/llama.swiftui/llama.cpp.swift/LibLlama.swift
View file

@ -164,13 +164,21 @@ actor LlamaContext {
private func token_to_piece(token: llama_token) -> String {
let result = UnsafeMutablePointer<Int8>.allocate(capacity: 8)
result.initialize(repeating: Int8(0), count: 8)
defer {
result.deallocate()
}
let nTokens = llama_token_to_piece(model, token, result, 8)
let _ = llama_token_to_piece(model, token, result, 8)
let resultStr = String(cString: result)
result.deallocate()
return resultStr
if nTokens < 0 {
let newResult = UnsafeMutablePointer<Int8>.allocate(capacity: Int(-nTokens))
newResult.initialize(repeating: Int8(0), count: Int(-nTokens))
defer {
newResult.deallocate()
}
_ = llama_token_to_piece(model, token, newResult, -nTokens)
return String(cString: newResult)
} else {
return String(cString: result)
}
}
}