lite-mamba
A minimal, pure-TensorFlow implementation of Mamba with a multi-dilated causal depthwise conv front-end. No custom C++ or Triton kernels needed; works seamlessly on CPU, GPU, or TPU with standard TensorFlow ops.
📚 READ THE FULL DOCUMENTATION HERE 📚
Contains architecture details, API references, streaming inference guides, and overviews of the multi-branch variants (TFPTCNMamba, TFSTCNMamba, TFDPWCMamba).
Quick Start
Install
pip install lite-mamba
Basic Usage
from lite_mamba import TFPTCNMamba
import tensorflow as tf
x = tf.random.normal((2, 128, 512)) # (batch, seq, d_model)
m = TFPTCNMamba(d_model=512, d_conv=3, conv_dilations=(1, 2, 4, 8))
y = m(x)
print(y.shape) # (2, 128, 512)
Changelog
See CHANGELOG.md. 1.2.0 fixes a save_weights() failure
(Unable to synchronously create dataset (name already exists)) that affected every
model containing a Mamba block on Keras 2 / TF ≤ 2.15.
License
Apache-2.0
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file lite_mamba-1.2.0.tar.gz.
File metadata
- Download URL: lite_mamba-1.2.0.tar.gz
- Upload date:
- Size: 12.9 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/7.0.0 CPython/3.11.15
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
3e3223c546b272fde3f1c12e6f540aced9ccef7d1b1e926b36edc867c32b9c23
|
|
| MD5 |
abffc9d0a287f62a3bbf83d7623b3c01
|
|
| BLAKE2b-256 |
547d16dab07b8f1e7eba1c19b38fa68214bb54109ecd258b9e55ffeba128e4db
|
File details
Details for the file lite_mamba-1.2.0-py3-none-any.whl.
File metadata
- Download URL: lite_mamba-1.2.0-py3-none-any.whl
- Upload date:
- Size: 8.4 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/7.0.0 CPython/3.11.15
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f34e05074a7fc6e9f3627e4d43b8ccc0a31b6c0bd0060d34fbbabd86e09d7942
|
|
| MD5 |
e2a69e1861a8b8248cf1ad585918eb6c
|
|
| BLAKE2b-256 |
87460f8bba90e1ad7c116f36ea0adc826e81892318d023155b0690267ed035cd
|
