Metadata-Version: 2.4
Name: titans-pytorch
Version: 0.1.31
Summary: Titans
Project-URL: Homepage, https://pypi.org/project/titans-pytorch/
Project-URL: Repository, https://github.com/lucidrains/titans-pytorch
Author-email: Phil Wang <lucidrains@gmail.com>
License: MIT License
        
        Copyright (c) 2025 Phil Wang
        
        Permission is hereby granted, free of charge, to any person obtaining a copy
        of this software and associated documentation files (the "Software"), to deal
        in the Software without restriction, including without limitation the rights
        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
        copies of the Software, and to permit persons to whom the Software is
        furnished to do so, subject to the following conditions:
        
        The above copyright notice and this permission notice shall be included in all
        copies or substantial portions of the Software.
        
        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
        SOFTWARE.
License-File: LICENSE
Keywords: artificial intelligence,deep learning,linear attention,neural memory module,test time training
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3.9
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.9
Requires-Dist: accelerated-scan>=0.2.0
Requires-Dist: axial-positional-embedding>=0.3.10
Requires-Dist: einops>=0.8.0
Requires-Dist: einx>=0.3.0
Requires-Dist: hyper-connections>=0.1.9
Requires-Dist: ninja
Requires-Dist: rotary-embedding-torch
Requires-Dist: tensordict
Requires-Dist: torch>=2.2
Requires-Dist: tqdm
Requires-Dist: x-transformers
Provides-Extra: examples
Requires-Dist: adam-atan2-pytorch>=0.1.18; extra == 'examples'
Requires-Dist: wandb; extra == 'examples'
Provides-Extra: test
Requires-Dist: pytest; extra == 'test'
Description-Content-Type: text/markdown

<img src="./fig2.png" width="400px"></img>

<img src="./fig1.png" width="400px"></img>

## Titans - Pytorch (wip)

Unofficial implementation of [Titans](https://arxiv.org/abs/2501.00663) in Pytorch. Will also contain some explorations into architectures beyond their simple 1-4 layer MLP for the neural memory module, if it works well to any degree.

## Appreciation

- [Eryk](https://github.com/sentialx) for sharing his early experimental results with me, positive for 2 layer MLP

## Install

```bash
$ pip install titans-pytorch
```

## Usage

```python
import torch
from titans_pytorch import NeuralMemory

mem = NeuralMemory(
    dim = 384,
    chunk_size = 64 # set to smaller chunk size for better perf on smaller sequence lengths (but more memory usage)
).cuda()

seq = torch.randn(2, 1024, 384).cuda()
retrieved = mem(seq)

assert seq.shape == retrieved.shape
```

A transformer with the `MAC` configuration can be used as

```python
import torch
from titans_pytorch import MemoryAsContextTransformer

transformer = MemoryAsContextTransformer(
    num_tokens = 256,
    dim = 256,
    depth = 2,
    segment_len = 128,              # local attention window size
    num_persist_mem_tokens = 4,
    num_longterm_mem_tokens = 16,
)

token_ids = torch.randint(0, 256, (1, 1023))

loss = transformer(token_ids, return_loss = True) # (1, 1023, 256)
loss.backward()

# after much training

sampled = transformer.sample(token_ids[:, :4], 512)
```

## Experiments

```bash
$ pip install .[examples]
```

Then modify `train_mac.py` and run it to query nature

```bash
$ python train_mac.py
```

## Citations

```bibtex
@inproceedings{Behrouz2024TitansLT,
    title   = {Titans: Learning to Memorize at Test Time},
    author  = {Ali Behrouz and Peilin Zhong and Vahab S. Mirrokni},
    year    = {2024},
    url     = {https://api.semanticscholar.org/CorpusID:275212078}
}
```

```bibtex
@article{Sun2024LearningT,
    title   = {Learning to (Learn at Test Time): RNNs with Expressive Hidden States},
    author  = {Yu Sun and Xinhao Li and Karan Dalal and Jiarui Xu and Arjun Vikram and Genghan Zhang and Yann Dubois and Xinlei Chen and Xiaolong Wang and Oluwasanmi Koyejo and Tatsunori Hashimoto and Carlos Guestrin},
    journal = {ArXiv},
    year    = {2024},
    volume  = {abs/2407.04620},
    url     = {https://api.semanticscholar.org/CorpusID:271039606}
}
```

```bibtex
@inproceedings{Yang2024GatedDN,
    title   = {Gated Delta Networks: Improving Mamba2 with Delta Rule},
    author  = {Songlin Yang and Jan Kautz and Ali Hatamizadeh},
    year    = {2024},
    url     = {https://api.semanticscholar.org/CorpusID:274598177}
}
```

```bibtex
@inproceedings{Nguyen2024TurningUT,
    title   = {Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputs},
    author  = {Minh Nguyen and Andrew Baker and Clement Neo and Allen Roush and Andreas Kirsch and Ravid Shwartz-Ziv},
    year    = {2024},
    url     = {https://api.semanticscholar.org/CorpusID:270870613}
}
```

```bibtex
@article{Zhu2024HyperConnections,
    title   = {Hyper-Connections},
    author  = {Defa Zhu and Hongzhi Huang and Zihao Huang and Yutao Zeng and Yunyao Mao and Banggu Wu and Qiyang Min and Xun Zhou},
    journal = {ArXiv},
    year    = {2024},
    volume  = {abs/2409.19606},
    url     = {https://api.semanticscholar.org/CorpusID:272987528}
}
```

```bibtex
@article{Zhou2024ValueRL,
    title   = {Value Residual Learning For Alleviating Attention Concentration In Transformers},
    author  = {Zhanchao Zhou and Tianyi Wu and Zhiyun Jiang and Zhenzhong Lan},
    journal = {ArXiv},
    year    = {2024},
    volume  = {abs/2410.17897},
    url     = {https://api.semanticscholar.org/CorpusID:273532030}
}
```

```bibtex
@software{Kyrylov_Accelerated_Scan_2024,
    author  = {Kyrylov, Volodymyr},
    doi     = {10.5281/zenodo.10600962},
    title   = {Accelerated Scan},
    version = {0.1.2},
    year    = {2024}
}
```

```bibtex
@misc{wang2025testtimeregressionunifyingframework,
    title   = {Test-time regression: a unifying framework for designing sequence models with associative memory},
    author  = {Ke Alexander Wang and Jiaxin Shi and Emily B. Fox},
    year    = {2025},
    eprint  = {2501.12352},
    archivePrefix = {arXiv},
    primaryClass = {cs.LG},
    url = {https://arxiv.org/abs/2501.12352},
}
```
