[BUG] NULL pointer access for text interface

[fbesoainp]

Issue Description

I’m encountering a ValueError: NULL pointer access error when calling create_completion or create_chat_completion in the NexaTextInference interface. This issue started unexpectedly, despite previously working fine. Restarting my computer didn’t resolve the issue.

Code sample:

from threading import Lock
from nexa_sdk import NexaTextInference, CreateChatCompletionResponse, CreateChatCompletionStreamResponse
from typing import Iterator

class NexaaiIntegration:
    _instance: NexaaiIntegration | None = None
    _lock: Lock = Lock()
    inference: NexaTextInference | None = None

    def __new__(cls, model_path="llama3-uncensored") -> NexaaiIntegration:
        with cls._lock:
            if cls._instance is None:
                cls._instance = super(NexaaiIntegration, cls).__new__(cls)
                cls._instance.inference = NexaTextInference(
                    model_path=model_path,
                    local_path=None,
                    stop_words=[],
                    temperature=0.5,
                    max_new_tokens=100,
                    top_k=20,
                    top_p=0.9,
                    profiling=True,
                    embedding=True
                )
                cls._instance.inference.create_embedding("You are an AI assistant for a software engineering team.")

            return cls._instance

    def create_text_completion(
        self, message: str
    ) -> CreateChatCompletionResponse | Iterator[CreateChatCompletionStreamResponse]:
        """Returns a completion for a chat message."""
        return self._instance.inference.create_completion(message)


nexa = NexaaiIntegration()
completion = nexa.create_text_completion("Hello, how are you?")
print(completion)

Steps to Reproduce

Run this code sample:

from threading import Lock
from nexa_sdk import NexaTextInference, CreateChatCompletionResponse, CreateChatCompletionStreamResponse
from typing import Iterator

class NexaaiIntegration:
    _instance: NexaaiIntegration | None = None
    _lock: Lock = Lock()
    inference: NexaTextInference | None = None

    def __new__(cls, model_path="llama3-uncensored") -> NexaaiIntegration:
        with cls._lock:
            if cls._instance is None:
                cls._instance = super(NexaaiIntegration, cls).__new__(cls)
                cls._instance.inference = NexaTextInference(
                    model_path=model_path,
                    local_path=None,
                    stop_words=[],
                    temperature=0.5,
                    max_new_tokens=100,
                    top_k=20,
                    top_p=0.9,
                    profiling=True,
                    embedding=True
                )
                cls._instance.inference.create_embedding("You are an AI assistant for a software engineering team.")

            return cls._instance

    def create_text_completion(
        self, message: str
    ) -> CreateChatCompletionResponse | Iterator[CreateChatCompletionStreamResponse]:
        """Returns a completion for a chat message."""
        return self._instance.inference.create_completion(message)


nexa = NexaaiIntegration()
completion = nexa.create_text_completion("Hello, how are you?")
print(completion)

OS

macOS

Python Version

3.11.7

Nexa SDK Version

0.0.9.0

GPU (if using one)

Apple M1 Pro