xcxhy

@zhuohan123 Hi, I use the vicuna_7b_1.3 to inference. My input text is very long, more than 1500 tokens, but I limited the maximum length to 2048, but after entering multiple...

@NLP-ZY Agree. I notice is that neither train_dumm.py nor train_prompts.py save the checkpoint.

@tpoisonooo thanks for your response, l have tried the llama.cpp before, and it can indeed accelerate on the cpu, but the inference speed is not significantly improved compared to the...

@tpoisonooo Thank you for response, but I'm still stuck at the convert onnx part. Sorry, I pulled your brank, but don't know how to convert onnx. I tried both onnx.export...

@tpoisonooo Thanks your response. I have studied carefully for many days, and basically got through the process between them, but now I am converting pytorch to ONNX, there will be...

@kebijuelun Hi, your answer is very useful. Is the parameter modification you mentioned a piece of V100? If I have 8 pieces of 3090, can I fine-tune the FlanT5XL model...

@kebijuelun thank to your response, if I use the 3090 can use the same type "bf16", the blip2_t5.py I will need to change parameters, like"text_ouput"?

@kebijuelun Hello, I want to ask again, did you fine-tune with eval_okvqa_zeroshot_flant5xl.sh? Do you have a single card or multiple cards, and how much memory do you have? I use...

Thank to your response, I will try it later. 魇 ***@***.***   ------------------ 原始邮件 ------------------ 发件人: "salesforce/LAVIS" ***@***.***>; 发送时间: 2023年3月14日(星期二) 下午2:16 ***@***.***>; ***@***.******@***.***>; 主题: Re: [salesforce/LAVIS] How to use your own dataset to train...

@dxli94 Thank you for your response, just went to test, the problem is solved. I have some code details, how long is the longest BLIP2 encoded text?