DeepEP icon indicating copy to clipboard operation
DeepEP copied to clipboard

DeepEP: an efficient expert-parallel communication library

Results 175 DeepEP issues
Sort by recently updated
recently updated
newest added

As shown in the figure, after assigning different num_tokens parameters to each RANK and running test_low_latency.py, the process gets stuck. What methods can be used to profile the impact of...

I'd like to know where **_synchronization across all ranks_** is required for both dispatch and combine operations, using the following code that calls low_latency_dispatch and low_latency_combine as an example. Specifically:...

test with 4 rank, 8 rank, 24 rank all pass. code is messy and I will refactor later, and also will try to improve a bit of performance if possible....

code is roughly like this; I will work on related things more which will also verify this PR more EDIT: it works well on the target hardware; I will try...

Hi teams, I would like to deploy and run DeepEP on an L20 device. However, I noticed that your repository mentions that NVLink-enabled devices are required when using NVSHMEM. The...

I meet a problem: this build ok: NVSHMEM_DIR=/home/wb/nvshmem python setup.py build #python tests/test_intranode.py Traceback (most recent call last): File "/home/wb/DeepEP/tests/test_intranode.py", line 7, in import deep_ep File "/home/wb/miniconda3/lib/python3.9/site-packages/deep_ep-1.0.0+d5ca449-py3.9-linux-x86_64.egg/deep_ep/__init__.py", line 3, in...

error when run test_low_latency.py: Allocating buffer size: 2115.111296 MB ... /root/nvshmem_src/src/host/init/init.cu:nvshmemi_check_state_and_init:1077: nvshmem API called before nvshmem_init /root/nvshmem_src/src/host/init/init.cu:nvshmemi_check_state_and_init:1077: nvshmem API called before nvshmem_init /root/nvshmem_src/src/host/init/init.cu:nvshmemi_check_state_and_init:1077: nvshmem API called before nvshmem_init /root/nvshmem_src/src/host/init/init.cu:nvshmemi_check_state_and_init:1077: nvshmem...

I am experiencing an issue with NVSHMEM failing to initialize due to transport errors. The error message indicates that NVSHMEM is unable to detect the system topology and cannot initialize...

When I ran test_internode.py in case of dual-port, the environment variable _**NVSHMEM_IBGDA_ENABLE_MULTI_PORT**_ is set to 1. Unfortunately, DeepEP crashed when create rdma team by calling nvshmem_team_split_strided(...) in the following code...