If I try to stop generate completion and cancel the coroutine job, CactusLM continues generating tokens until it's finished

[dmitriy-chernysh]

Am I doing something wrong? Is there any other way to stop the process of generation? Thanks

Library version: 1.0.2-beta
Android OS. No matter what Android version I use.

private suspend fun generateCompletionCancellable(
        messages: List<ChatMessage>,
        params: CactusCompletionParams,
        onToken: (String) -> Unit
    ): CactusCompletionResult? = suspendCancellableCoroutine { cont ->

        val workerJob = SupervisorJob()
        val workerScope = CoroutineScope(Dispatchers.IO + workerJob)

        workerScope.launch {
            try {
                cactusLM.generateCompletion(
                    messages = messages,
                    params = params,
                    onToken = { token, _ ->
                        Log.d("debug", "cactusLM token: $token")
                        if (!isActive) return@generateCompletion
                        onToken(token)
                    }
                )
                try {
                    cont.resume(result)
                } catch (_: Throwable) {
                }
            } catch (t: Throwable) {
                try {
                    cont.resumeWithException(t)
                } catch (_: Throwable) {
                }
            }
        }

        cont.invokeOnCancellation {
            workerJob.cancel()
            Log.d("debug", "cactusLM token: cancel the job")
        }
    }

From logs