fix: Alibaba Cloud full modal docking error

[shaohuzhang1]

fix: Alibaba Cloud full modal docking error --bug=1052232 --user=王孝刚 【模型设置】-大语言模型/语音识别模型 对接不上阿里云的全模态模型 https://www.tapd.cn/57709429/s/1655034

[shaohuzhang1]

There is one potential issue in the provided Python code:

The new_instance method does not handle other models besides 'qwen-omni-turbo'. If another model requires streaming to be enabled, this logic should also include that functionality.

Here is a revised version of the new_instance method with added support for additional models:

from typing import Dict

def new_instance(model_type, model_name, model_credential: Dict[str, object], **model_kwargs):
    optional_params = MaxKBBaseModel.filter_optional_params(model_kwargs)
    
    try:
        # Apply default stream parameter based on known supported models
        if model_name == 'qwen-omni-turbo' or model_name == 'another_supported_model':
            optional_params['streaming'] = True
        
        return BaiLianChatModel(
            model=model_name,
            openai_api_base=model_credential.get('api_base'),
            openai_access_token=model_credential.get('access_key'),
            prompt_history_size=model_credential.get('prompt_history_size', 1),
            max_seq_len=model_credential.get('max_output_length', 4096),
            temperature=model_credential.get('temperature', 0.7),
            top_p=model_credential.get('top_p', 0.95),
            history_num=model_credential.get('history_capacity', 2),
            **optional_params
        )
    except AttributeError as e:
        print(f"An error occurred while setting up the model instance: {e}")

In this updated function, I've included placeholder values (another_supported_model) where you would need to add any additional models that might require streaming. The function catches AttributeError to ensure robustness even if there's an unexpected configuration change within your application. This ensures all relevant parameters are correctly set when creating a BaiLianChatModel.

[shaohuzhang1]

The provided code appears to be part of a function that initializes an instance of AliyunBaiLianSpeechToText with various parameters. It checks and includes certain key settings based on the input model_kwargs. Here are some observations:

1. Conditional Streaming: The line if model_name == 'qwen-omni-turbo': optional_params['streaming'] = True ensures that streaming mode is enabled only when the model name is 'qwen-omni-turbo'. This is useful if there's specific behavior or integration requirements associated with this model.

2. Temperature Check: The existing checks for max_tokens and temperature ensure that they are set correctly before being passed to the AliyunBaiLianSpeechToText constructor, which seems appropriate.

3. Code Structure: The structure is neat and follows Python style guidelines. There's no apparent issue with readability or consistency in the code.

Summary of Changes:

No changes were suggested since the current implementation already addresses the expected use cases and best practices for setting up the parameters. If streaming isn't typically required across all instances of this speech-to-text service, it might make sense to remove the conditionals around enabling streaming.

If you need further modifications or optimizations, feel free to let me know!