Neo
> It seems the authors took it down when preparing to release EAGLE 3. The generating script exists on older commits such as: https://github.com/SafeAILab/EAGLE/tree/7e885b2ba9cf6d116cd041790f06143720821a68 > > Alternatively, I'm maintaining an...
I have the same questions as @junghye01 , did u solved these problems or anybody can help with this ,thanks
During the prefill phase of inference, if this parameter is set to False, all traffic will be forwarded through IB NIC 0, which will cause a bandwidth bottleneck for the...
@sphish Thanks for Ur Reply, I had setted these two environment variables before raising this issue, but there is still an imbalance in the network card traffic in normal mode....
I see, it means that when low_latency_mode = False, I need to set different 'NVSHMEM_HCA_LIST' for each GPU rank ? if I do like this, will the IB NICs become...
@u4lr451 hi, does this pr has any problems, and when will it can be merged into main, thanks
same error on 4 * 8 * H800, any progress with that @newfarmer thanks !
@ZJY0516 thanks for ur reply, I just modify text_to_image.py and run this script , it runs normally on H100 *8, but will crash on 5090, maybe vllm-omni is not supported...
@ZJY0516 @hsliuustc0106 Thanks, I think I'll try to get it running on the 5090 by myself