BossPi
BossPi
Thank you for your question. I'm sorry for the late reply. If you want to input a video for inference, you can organize your messages in the following way: ```...
Yes, it will same work if the video is a local .mp4 path.
Ernie processing video at fps:2 and max pixels:1196 * 28 * 28. We don't recommend to disable thinking mode. If you do want to disable thinking mode, you can process...
I'm glad to be of help to you.
您好，您可以参考如下方式传入本地图片： ``` messages = [ { "role": "user", "content": [ {"type": "text", "text": "What color clothes is the girl in the picture wearing?"}, {"type": "image_url", "image_url": {"url": "local_image.jpg"}}, ] },...
请问方便提供您的运行代码和具体的报错信息吗？
Thank you very much for your interest in and inquiry about our data distribution curation process. Regarding your question about how we calculate the token count for each data entry,...
We control the proportion of different types of data by adjusting the sampling ratio. Typically, we configure it based on the number of epochs for samples. For example, we sample...
您好，这个问题是由于transformers版本更新导致的，将transformers降级到4.53.0版本或者在预测时传入`use_cache=False`即可： ``` generated_ids = model.generate( inputs=inputs['input_ids'].to(device), **inputs, max_new_tokens=128, use_cache=False ) ```
不客气，如果还有其他问题，欢迎继续提问，我们会尽快为您解答