Sungjae Lee

I agree with the way to block duplicated information to be appeared in the merged file and to use the info from the first file found. I'm also relieved that...

I didn't think about a new option and I also think it is complex. How about just making users pick a single file to add the information on the `.swagger.yaml`...

Oh, I'm sorry, if I made you confused. I only used your external openapi configuration file option. Because I didn't know your internal openapi annotation. I made our Makefile as...

Yes, I would be willing to contribute a documentation update. Thank you.

Hello, I have similar issue to generate single swagger file from multiple proto files. I found out that there was an effort to generate single swagger output from multiple proto...

oh wow, thanks for the info. 👍 I was so confused with lots of issues, but finally I could generate swagger file that I want, thanks for your comment.

https://arxiv.org/pdf/2003.10555.pdf It seems that this paper also represents the inference FLOPs of `BERT-BASE` as 29 BFLOPs. so 21785M looked quite affordable for me.

Thank you @oyvindwe, I got a great help!

@jackzhou121 , I had same issues, and resolved them using following patch. This code casts the value into a double, ensuring that the compiler always chooses double operands for the...

I am experiencing the same issue. I tested with vLLM v0.4.1 and flash_attn v2.5.7. I also tested with the GQA and MHA models, both with and without Tensor Parallelism, and...