mamba
mamba copied to clipboard
state-spaces
There are three versions here, corresponding to mamba2.py, mamba2_simple.py, and ssd_minimal.py. The first two versions require the mamba_ssm library. What are the differences between them? Does this mean that the...
AMD ROCm 6.1.3 WSL2 Official PyTorch 2.1.2 https://rocm.blogs.amd.com/artificial-intelligence/mamba/README.html ROCm users on WSL2 HAVE to use this version of python. It the only officially supported one and its the only one...
Clarifying that "no build isolation" option might help with other problems, not only PyTorch versions.
Hey Tri and Albert, Digging through the code base, I have found a bug in the backward pass for the gradient calculations in the function `_chunk_scan_bwd_ddAcs_stable_kernel`. I think this solves...
/data/miniconda3/envs/mamba/bin/python /data/mamba/benchmarks/benchmark_generation_mamba_simple.py Loading model state-spaces Traceback (most recent call last): File "/data/miniconda3/envs/mamba/lib/python3.10/site-packages/urllib3/connection.py", line 174, in _new_conn conn = connection.create_connection( File "/data/miniconda3/envs/mamba/lib/python3.10/site-packages/urllib3/util/connection.py", line 95, in create_connection raise err File "/data/miniconda3/envs/mamba/lib/python3.10/site-packages/urllib3/util/connection.py", line...
>>> from mamba_ssm import Mamba Traceback (most recent call last): File "", line 1, in File "/home/deng/miniforge-pypy3/envs/mamba/lib/python3.10/site-packages/mamba_ssm/__init__.py", line 3, in from mamba_ssm.ops.selective_scan_interface import selective_scan_fn, mamba_inner_fn File "/home/deng/miniforge-pypy3/envs/mamba/lib/python3.10/site-packages/mamba_ssm/ops/selective_scan_interface.py", line 16, in...
Hi author! Thanks for your brilliant work first. I try to calculate the flops of mamba through calculate_flops from calflops library. I am wondering if the efficiency of selective scan...
Hi, I'm experiencing an issue with `clip_grad_norm_` and loss values while training Mamba2. After training for some time, the gradient norm starts to rapidly increase to infinity. If training continues,...
Hello Thank you for the awesome work. I installed mamba-ssm and used it in my projects and it worked perfectly fine. when I use hydra in my python script I...
I have a 4-stage network, and considering that each stage has a different number of tokens, I want to set different sizes for d_state, e.g., [256, 128, 64, 32]. However,...
