Skip to content

Commit

Permalink
swift : fix token_to_piece implementation (#4278)
Browse files Browse the repository at this point in the history
* Fix token_to_piece implementation in Swift

* Fix errors
  • Loading branch information
ensan-hcl authored Dec 1, 2023
1 parent 511f52c commit b220222
Show file tree
Hide file tree
Showing 2 changed files with 19 additions and 15 deletions.
10 changes: 3 additions & 7 deletions examples/batched.swift/Sources/main.swift
Original file line number Diff line number Diff line change
Expand Up @@ -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))
}
Expand All @@ -259,5 +256,4 @@ private func token_to_piece(token: llama_token, buffer: inout [CChar]) -> String
buffer = []
return bufferString
}
return nil
}
24 changes: 16 additions & 8 deletions examples/llama.swiftui/llama.cpp.swift/LibLlama.swift
Original file line number Diff line number Diff line change
Expand Up @@ -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)

let _ = llama_token_to_piece(model, token, result, 8)

let resultStr = String(cString: result)

result.deallocate()

return resultStr
defer {
result.deallocate()
}
let nTokens = llama_token_to_piece(model, token, result, 8)

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)
}
}
}

0 comments on commit b220222

Please sign in to comment.