mamba-ssm

library2.3.2.post1pypypiunverified

Mamba is a state-space model architecture designed for efficient sequence modeling, offering linear-time inference and parallelizable training. Currently at version 2.3.1, it requires Python >=3.9 and is under active development with frequent releases focused on ROCm and CUDA compatibility.

ai-ml

pip install mamba-ssm

github ↗package ↗

INSTALL

IMPORT

SIG · MAMBA-SSM

mamba-ssm

ai-mlpythonv2.3.2.post1

Install

—

Import

—

Disk

—

Pass rate

0/ 10

Env Coverage0 / 10

glibc

3.9–3.13

musl

3.9–3.13

Install & Compatibility

Where this runs

tested against v? · pip install

Install × environment matrix

Each cell = how many times install + import succeeded across repeated harness runs. Partial = flaky.
glibc = Debian/Ubuntu slim · musl = Alpine Linux

musl

py 3.10–3.910 runs

build_error

glibc

py 3.10–3.910 runs

build_error

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

Mamba

✓ from mamba_ssm import Mamba

✗ import mamba_ssm

Direct import of class is required; the top-level module does not expose the class automatically.

Mamba2

✓ from mamba_ssm import Mamba2

Mamba2 is a variant and must be imported similarly.

Instantiate a Mamba model and run a forward pass on GPU with float16 precision.

import torch from mamba_ssm import Mamba batch, seq_len, dim = 2, 128, 16 model = Mamba( d_model=dim, d_state=16, d_conv=4, expand_factor=2, dt_rank='auto', bias=False, conv_bias=True, pscan=False, device='cuda', dtype=torch.float16 ) x = torch.randn(batch, seq_len, dim, device='cuda', dtype=torch.float16) y = model(x) print(y.shape) # Expected: (2, 128, 16)

Debug

Known issues

breakingMamba v2.x requires PyTorch >=2.0 and CUDA 11.8+ for GPU support. Older PyTorch versions will fail with missing ops.

fix

Upgrade PyTorch to 2.0+ and ensure CUDA toolkit 11.8+ is available.

affects: >=2.0.0

breakingThe causal-conv1d dependency is a separate package that may have ABI incompatibilities with different PyTorch versions. Mixing builds can cause silent incorrect results or crashes.

fix

Install causal-conv1d from the same source (PyPI with matching CUDA version) or build from source using the same PyTorch build.

affects: all

deprecatedThe 'pscan' parameter in Mamba is deprecated since v2.0 and will be removed in a future release. Setting pscan=True may lead to undefined behavior.

fix

Remove the 'pscan' argument or set it to False (default).

affects: >=2.0.0

gotchaThe Mamba model expects input shape (batch, seq_len, d_model). Transposing or using (seq_len, batch, d_model) will not raise an error but will produce incorrect outputs due to dimension mismatch.

fix

Ensure input shape is (B, L, D). Use x = x.transpose(0,1) if you have (L, B, D).

affects: all

gotchaWhen using torch.compile, the custom CUDA kernels may not be compatible. Expect failures or performance degradation.

fix

Avoid torch.compile with Mamba layers, or test thoroughly.

affects: all

Upgrade

Version history

2.3.2.post1latest on PyPI · released May 9, 2026

Audit

Dependencies

torchrequiredPyTorch is a core dependency for tensor operations and CUDA support.

causal-conv1doptionalOptional but recommended for faster causal convolution layers.

packagingrequiredUsed internally for version checks.

einopsrequiredRequired for rearranging tensors.

Agent activity

12 hits · last 30 days

node

Resources

githubhttps://github.com/state-spaces/mamba ↗

packagehttps://pypi.org/project/mamba-ssm/ ↗