FL77N
FL77N
> y (cx, cy, w, h) and (x1, y1, x2, y2) can b Em~thank you, I mean I should use the format (x1, y1, x2, y2) to compute l1 loss...
> yep~ thanks
> _No description provided._ hello，have you exported onnx successfully?
> 可能是某些OP无法推断出shape导致还是存在-1，请问你的部署场景是？这样的 您好，您解决了这个问题了吗？我也有同样的问题。
> 具体是执行什么指令时遇到的？ 下不了，然后用 debug 发现的 !you-get --itag=137 --debug "https://www.youtube.com/watch?v=FFoNw-dYQf0&list=RDFFoNw-dYQf0&start_radio=1&rv=FFoNw-dYQf0&t=2"
@jklj077 Hi, when I use transformers to sft qwen2 57b moe on 32xA100 80G with input length 2048, it is oom. is there something wrong with my usage?
> For reference, full parameter finetuning for Qwen2-57B-A14B should be possible with 2 * 8 * 80GB GPUs with 4K sequence length (estimated minimum). However, you should enable an mixture...