Li Bo

We did change the structure of HF's dataset into parquets (inside it, it's a Dataframe object) as suggested by HF staff. So this dataset would be visible to viewers. It...

Hi sorry we've changed to the serving/hosting logic from original `fastchat` (since it involves so many codes and files, we wish to make it easier and quicker to host new...

Yes, unless you set the batch_size to 1, you can not use image and video dataset together because they have different shapes. Otter is designed to support multi-modal in-context instruction...

> thank you, I want to know if I want to reproduce the result of image-text model, which dataset should I use? hi do you mean what result? For short...

I guess you should use this file as Deepspeed training config? https://github.com/Luodian/Otter/blob/main/pipeline/accelerate_configs/accelerate_config_zero2.yaml

hi Vishaal, may I know if you are using one GPU with 40G to train? Can you decrease it to `batch_size=1` or try with`luodian/OTTER-MPT1B-RPJama-Init`?

> hi Vishaal, may I know if you are using one GPU with 40G to train? Can you decrease it to `batch_size=1` or try with`luodian/OTTER-MPT1B-RPJama-Init`? I am not sure if...

We dont have a promising result for lora finetuning. We tried finetune perceiver + cross_x_attn + lora LLM. But dont get better results than finetuning perceiver + cross_x_attn. If you...

If loaded with lora LLM, you could see relevant logs to show how many params are LoRAed, etc.

oh that depends on how you set it in training, if you choose load the `SD` and `GSD` 's two images as in-context examples. The prompt should be `User: What's...