# IBM Analog Hardware Acceleration Kit
[](https://aihwkit.readthedocs.io/en/latest/?badge=latest)
[](https://travis-ci.com/IBM/aihwkit)
[](https://arxiv.org/abs/2104.02184)
## Description
_IBM Analog Hardware Acceleration Kit_ is an open source Python toolkit for
exploring and using the capabilities of in-memory computing devices in the
context of artificial intelligence.
> :warning: This library is currently in beta and under active development.
> Please be mindful of potential issues and keep an eye for improvements,
> new features and bug fixes in upcoming versions.
The toolkit consists of two main components:
### Pytorch integration
A series of primitives and features that allow using the toolkit within
[`PyTorch`]:
* Analog neural network modules (fully connected layer, 1d/2d/3d convolution
layers, LSTM layer, sequential container).
* Analog training using torch training workflow:
* Analog torch optimizers (SGD).
* Analog in-situ training using customizable device models and algorithms
(Tiki-Taka).
* Analog inference using torch inference workflow:
* State-of-the-art statistical model of a phase-change memory (PCM) array
calibrated on hardware measurements from a 1 million PCM devices chip.
* Hardware-aware training with hardware non-idealities and noise
included in the forward pass to make the trained models more
robust during inference on Analog hardware.
### Analog devices simulator
A high-performant (CUDA-capable) C++ simulator that allows for
simulating a wide range of analog devices and crossbar configurations
by using abstract functional models of material characteristics with
adjustable parameters. Features include:
* Forward pass output-referred noise and device fluctuations, as well
as adjustable ADC and DAC discretization and bounds
* Stochastic update pulse trains for rows and columns with finite
weight update size per pulse coincidence
* Device-to-device systematic variations, cycle-to-cycle noise and
adjustable asymmetry during analog update
* Adjustable device behavior for exploration of material specifications for
training and inference
* State-of-the-art dynamic input scaling, bound management, and update
management schemes
### Other features
Along with the two main components, the toolkit includes other
functionalities such as:
* A library of device presets that are calibrated to real hardware data and
based on models in the literature, along with a configuration that specifies a particular device and optimizer choice.
* A module for executing high-level use cases ("experiments"), such as neural
network training with minimal code overhead.
* A utility to automatically convert a downloaded model (e.g., pre-trained) to its equivalent Analog
model by replacing all linear/conv layers to Analog layers (e.g., for convenient hardware-aware training).
* Integration with the [AIHW Composer] platform, a no-code web experience that allows executing
experiments in the cloud.
## How to cite?
In case you are using the _IBM Analog Hardware Acceleration Kit_ for
your research, please cite the AICAS21 paper that describes the toolkit:
> Malte J. Rasch, Diego Moreda, Tayfun Gokmen, Manuel Le Gallo, Fabio Carta,
> Cindy Goldberg, Kaoutar El Maghraoui, Abu Sebastian, Vijay Narayanan.
> ["A flexible and fast PyTorch toolkit for simulating training and inference on
> analog crossbar arrays"](https://ieeexplore.ieee.org/abstract/document/9458494) (2021 IEEE 3rd International Conference on Artificial Intelligence Circuits and Systems)
>
## Usage
### Training example
```python
from torch import Tensor
from torch.nn.functional import mse_loss
# Import the aihwkit constructs.
from aihwkit.nn import AnalogLinear
from aihwkit.optim import AnalogSGD
x = Tensor([[0.1, 0.2, 0.4, 0.3], [0.2, 0.1, 0.1, 0.3]])
y = Tensor([[1.0, 0.5], [0.7, 0.3]])
# Define a network using a single Analog layer.
model = AnalogLinear(4, 2)
# Use the analog-aware stochastic gradient descent optimizer.
opt = AnalogSGD(model.parameters(), lr=0.1)
opt.regroup_param_groups(model)
# Train the network.
for epoch in range(10):
pred = model(x)
loss = mse_loss(pred, y)
loss.backward()
opt.step()
print('Loss error: {:.16f}'.format(loss))
```
You can find more examples in the [`examples/`] folder of the project, and
more information about the library in the [documentation]. Please note that
the examples have some additional dependencies - you can install them via
`pip install -r requirements-examples.txt`.
You can find interactive notebooks and tutorials in the [`notebooks/`] directory.
### Further reading
We also recommend to take a look at the tutorial article that
describes the usage of the toolkit that can be found here:
> Manuel Le Gallo, Corey Lammie, Julian Buechel, Fabio Carta, Omobayode Fagbohungbe,
> Charles Mackin, Hsinyu Tsai, Vijay Narayanan, Abu Sebastian, Kaoutar El Maghraoui,
> Malte J. Rasch.
> ["Using the IBM Analog In-Memory Hardware Acceleration Kit for Neural Network Training and Inference"](https://doi.org/10.1063/5.0168089)
> (APL Machine Learning Journal:1(4) 2023)
>
## What is Analog AI?
In traditional hardware architecture, computation and memory are siloed in
different locations. Information is moved back and forth between computation
and memory units every time an operation is performed, creating a limitation
called the [von Neumann bottleneck].
Analog AI delivers radical performance improvements by combining compute and
memory in a single device, eliminating the von Neumann bottleneck. By leveraging
the physical properties of memory devices, computation happens at the same place
where the data is stored. Such in-memory computing hardware increases the speed
and energy efficiency needed for next-generation AI workloads.
## What is an in-memory computing chip?
An in-memory computing chip typically consists of multiple arrays of memory
devices that communicate with each other. Many types of memory devices such as
[phase-change memory] (PCM), [resistive random-access memory] (RRAM), and
[Flash memory] can be used for in-memory computing.
Memory devices have the ability to store synaptic weights in their analog
charge (Flash) or conductance (PCM, RRAM) state. When these devices are arranged
in a crossbar configuration, it allows to perform an analog matrix-vector
multiplication in a single time step, exploiting the advantages of analog
storage capability and [Kirchhoff’s circuits laws]. You can learn more about
it in our [online demo].
In deep learning, data propagation through multiple layers of a neural network
involves a sequence of matrix multiplications, as each layer can be represented
as a matrix of synaptic weights. The devices are arranged in multiple crossbar
arrays, creating an artificial neural network where all matrix multiplications
are performed in-place in an analog manner. This structure allows to run deep
learning models at reduced energy consumption.
## Awards and Media Mentions
* IBM Research blog: [Open-sourcing analog AI simulation]: <https://research.ibm.com/blog/analog-ai-for-efficient-computing>
* We are proud to share that the AIHWKIT and the companion cloud composer received the IEEE OPEN SOURCE SCIENCE [award] in 2023.
<img width="817" alt="IEEE Award" src="https://github.com/IBM/aihwkit/assets/7916630/bd3347fe-49c0-4aa2-ba6e-455b0e2a91ce">
## Installation
### Installing from PyPI
The preferred way to install this package is by using the
[Python package index]:
```shell
pip install aihwkit
```
### Conda-based Installation
There is a conda package for aihwkit available in conda-forge. It can be installed in a conda environment running on a Linux or WSL in a Windows system.
* CPU
```shell
conda install -c conda-forge aihwkit
```
* GPU
```shell
conda install -c conda-forge aihwkit-gpu
```
If you encounter any issues during download or want to compile the package
for your environment, please take a look at the [advanced installation] guide.
That section describes the additional libraries and tools required for
compiling the sources using a build system based on `cmake`.
### Docker Installation
For GPU support, you can also build a docker container following the [CUDA Dockerfile instructions].
You can then run a GPU enabled docker container using the follwing command from your peoject dircetory
```shell
docker run --rm -it --gpus all -v $(pwd):$HOME --name aihwkit aihwkit:cuda bash
```
## Authors
IBM Research has developed IBM Analog Hardware Acceleration Kit,
with Malte Rasch, Diego Moreda, Fabio Carta, Julian Büchel, Corey Lammie, Charles Mackin, Kim Tran, Tayfun Gokmen, Manuel Le Gallo-Bourdeau, and Kaoutar El Maghraoui
as the initial core authors, along with many [contributors].
You can contact us by opening a new issue in the repository or alternatively
at the ``aihwkit@us.ibm.com`` email address.
## License
This project is licensed under [Apache License 2.0].
[Apache License 2.0]: LICENSE.txt
[Python package index]: https://pypi.org/project/aihwkit
[`PyTorch`]: https://pytorch.org/
[`examples/`]: examples/
[`notebooks/`]: notebooks/
[documentation]: https://aihwkit.readthedocs.io/
[contributors]: https://github.com/IBM/aihwkit/graphs/contributors
[advanced installation]: https://aihwkit.readthedocs.io/en/latest/advanced_install.html
[von Neumann bottleneck]: https://en.wikipedia.org/wiki/Von_Neumann_architecture#Von_Neumann_bottleneck
[phase-change memory]: https://en.wikipedia.org/wiki/Phase-change_memory
[resistive random-access memory]: https://en.wikipedia.org/wiki/Resistive_random-access_memory
[Flash memory]: https://en.wikipedia.org/wiki/Flash_memory
[Kirchhoff’s circuits laws]: https://en.wikipedia.org/wiki/Kirchhoff%27s_circuit_laws
[online demo]: https://analog-ai-demo.mybluemix.net/
[AIHW Composer]: https://aihw-composer.draco.res.ibm.com
[award]: https://conferences.computer.org/services/2023/awards/
[CUDA Dockerfile instructions]: https://github.com/IBM/aihwkit/blob/master/docs/source/advanced_install.rst#cuda-enabled-docker-image
Raw data
{
"_id": null,
"home_page": "https://github.com/IBM/aihwkit",
"name": "aihwkit",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.7",
"maintainer_email": null,
"keywords": "ai, analog, rpu, torch, memristor, pcm, reram, crossbar, in-memory, nvm, non-von-neumann, non-volatile memory, phase-change material",
"author": "IBM Research",
"author_email": "aihwkit@us.ibm.com",
"download_url": "https://files.pythonhosted.org/packages/7b/ed/385adb141f3691000c33adb3d76c80e1fe213d3e991b91054066e4f698b6/aihwkit-0.9.2.tar.gz",
"platform": null,
"description": "# IBM Analog Hardware Acceleration Kit\n\n\n[](https://aihwkit.readthedocs.io/en/latest/?badge=latest)\n[](https://travis-ci.com/IBM/aihwkit)\n\n[](https://arxiv.org/abs/2104.02184)\n\n## Description\n\n_IBM Analog Hardware Acceleration Kit_ is an open source Python toolkit for\nexploring and using the capabilities of in-memory computing devices in the\ncontext of artificial intelligence.\n\n> :warning: This library is currently in beta and under active development.\n> Please be mindful of potential issues and keep an eye for improvements,\n> new features and bug fixes in upcoming versions.\n\nThe toolkit consists of two main components:\n\n### Pytorch integration\n\nA series of primitives and features that allow using the toolkit within\n[`PyTorch`]:\n\n* Analog neural network modules (fully connected layer, 1d/2d/3d convolution\n layers, LSTM layer, sequential container).\n* Analog training using torch training workflow:\n * Analog torch optimizers (SGD).\n * Analog in-situ training using customizable device models and algorithms\n (Tiki-Taka).\n* Analog inference using torch inference workflow:\n * State-of-the-art statistical model of a phase-change memory (PCM) array\n calibrated on hardware measurements from a 1 million PCM devices chip.\n * Hardware-aware training with hardware non-idealities and noise\n included in the forward pass to make the trained models more\n robust during inference on Analog hardware.\n\n### Analog devices simulator\n\nA high-performant (CUDA-capable) C++ simulator that allows for\nsimulating a wide range of analog devices and crossbar configurations\nby using abstract functional models of material characteristics with\nadjustable parameters. Features include:\n\n* Forward pass output-referred noise and device fluctuations, as well\n as adjustable ADC and DAC discretization and bounds\n* Stochastic update pulse trains for rows and columns with finite\n weight update size per pulse coincidence\n* Device-to-device systematic variations, cycle-to-cycle noise and\n adjustable asymmetry during analog update\n* Adjustable device behavior for exploration of material specifications for\n training and inference\n* State-of-the-art dynamic input scaling, bound management, and update\n management schemes\n\n### Other features\n\nAlong with the two main components, the toolkit includes other\nfunctionalities such as:\n\n* A library of device presets that are calibrated to real hardware data and\n based on models in the literature, along with a configuration that specifies a particular device and optimizer choice.\n* A module for executing high-level use cases (\"experiments\"), such as neural\n network training with minimal code overhead.\n* A utility to automatically convert a downloaded model (e.g., pre-trained) to its equivalent Analog\n model by replacing all linear/conv layers to Analog layers (e.g., for convenient hardware-aware training).\n* Integration with the [AIHW Composer] platform, a no-code web experience that allows executing\n experiments in the cloud.\n\n## How to cite?\n\nIn case you are using the _IBM Analog Hardware Acceleration Kit_ for\nyour research, please cite the AICAS21 paper that describes the toolkit:\n\n> Malte J. Rasch, Diego Moreda, Tayfun Gokmen, Manuel Le Gallo, Fabio Carta,\n> Cindy Goldberg, Kaoutar El Maghraoui, Abu Sebastian, Vijay Narayanan.\n> [\"A flexible and fast PyTorch toolkit for simulating training and inference on\n> analog crossbar arrays\"](https://ieeexplore.ieee.org/abstract/document/9458494) (2021 IEEE 3rd International Conference on Artificial Intelligence Circuits and Systems)\n>\n\n## Usage\n\n### Training example\n\n```python\nfrom torch import Tensor\nfrom torch.nn.functional import mse_loss\n\n# Import the aihwkit constructs.\nfrom aihwkit.nn import AnalogLinear\nfrom aihwkit.optim import AnalogSGD\n\nx = Tensor([[0.1, 0.2, 0.4, 0.3], [0.2, 0.1, 0.1, 0.3]])\ny = Tensor([[1.0, 0.5], [0.7, 0.3]])\n\n# Define a network using a single Analog layer.\nmodel = AnalogLinear(4, 2)\n\n# Use the analog-aware stochastic gradient descent optimizer.\nopt = AnalogSGD(model.parameters(), lr=0.1)\nopt.regroup_param_groups(model)\n\n# Train the network.\nfor epoch in range(10):\n pred = model(x)\n loss = mse_loss(pred, y)\n loss.backward()\n\n opt.step()\n print('Loss error: {:.16f}'.format(loss))\n```\n\nYou can find more examples in the [`examples/`] folder of the project, and\nmore information about the library in the [documentation]. Please note that\nthe examples have some additional dependencies - you can install them via\n`pip install -r requirements-examples.txt`.\nYou can find interactive notebooks and tutorials in the [`notebooks/`] directory.\n\n### Further reading\n\nWe also recommend to take a look at the tutorial article that\ndescribes the usage of the toolkit that can be found here:\n\n> Manuel Le Gallo, Corey Lammie, Julian Buechel, Fabio Carta, Omobayode Fagbohungbe,\n> Charles Mackin, Hsinyu Tsai, Vijay Narayanan, Abu Sebastian, Kaoutar El Maghraoui,\n> Malte J. Rasch.\n> [\"Using the IBM Analog In-Memory Hardware Acceleration Kit for Neural Network Training and Inference\"](https://doi.org/10.1063/5.0168089)\n> (APL Machine Learning Journal:1(4) 2023)\n>\n\n\n## What is Analog AI?\n\nIn traditional hardware architecture, computation and memory are siloed in\ndifferent locations. Information is moved back and forth between computation\nand memory units every time an operation is performed, creating a limitation\ncalled the [von Neumann bottleneck].\n\nAnalog AI delivers radical performance improvements by combining compute and\nmemory in a single device, eliminating the von Neumann bottleneck. By leveraging\nthe physical properties of memory devices, computation happens at the same place\nwhere the data is stored. Such in-memory computing hardware increases the speed\nand energy efficiency needed for next-generation AI workloads.\n\n## What is an in-memory computing chip?\n\nAn in-memory computing chip typically consists of multiple arrays of memory\ndevices that communicate with each other. Many types of memory devices such as\n[phase-change memory] (PCM), [resistive random-access memory] (RRAM), and\n[Flash memory] can be used for in-memory computing.\n\nMemory devices have the ability to store synaptic weights in their analog\ncharge (Flash) or conductance (PCM, RRAM) state. When these devices are arranged\nin a crossbar configuration, it allows to perform an analog matrix-vector\nmultiplication in a single time step, exploiting the advantages of analog\nstorage capability and [Kirchhoff\u2019s circuits laws]. You can learn more about\nit in our [online demo].\n\nIn deep learning, data propagation through multiple layers of a neural network\ninvolves a sequence of matrix multiplications, as each layer can be represented\nas a matrix of synaptic weights. The devices are arranged in multiple crossbar\narrays, creating an artificial neural network where all matrix multiplications\nare performed in-place in an analog manner. This structure allows to run deep\nlearning models at reduced energy consumption.\n\n## Awards and Media Mentions\n\n* IBM Research blog: [Open-sourcing analog AI simulation]: <https://research.ibm.com/blog/analog-ai-for-efficient-computing>\n* We are proud to share that the AIHWKIT and the companion cloud composer received the IEEE OPEN SOURCE SCIENCE [award] in 2023.\n <img width=\"817\" alt=\"IEEE Award\" src=\"https://github.com/IBM/aihwkit/assets/7916630/bd3347fe-49c0-4aa2-ba6e-455b0e2a91ce\">\n\n## Installation\n\n### Installing from PyPI\n\nThe preferred way to install this package is by using the\n[Python package index]:\n\n```shell\npip install aihwkit\n```\n\n### Conda-based Installation\nThere is a conda package for aihwkit available in conda-forge. It can be installed in a conda environment running on a Linux or WSL in a Windows system. \n\n* CPU\n\n ```shell\n conda install -c conda-forge aihwkit\n ```\n\n* GPU \n\n ```shell\n conda install -c conda-forge aihwkit-gpu\n ```\n\nIf you encounter any issues during download or want to compile the package\nfor your environment, please take a look at the [advanced installation] guide.\nThat section describes the additional libraries and tools required for\ncompiling the sources using a build system based on `cmake`.\n\n### Docker Installation \n\nFor GPU support, you can also build a docker container following the [CUDA Dockerfile instructions].\nYou can then run a GPU enabled docker container using the follwing command from your peoject dircetory\n\n```shell\ndocker run --rm -it --gpus all -v $(pwd):$HOME --name aihwkit aihwkit:cuda bash\n```\n\n## Authors\n\nIBM Research has developed IBM Analog Hardware Acceleration Kit,\nwith Malte Rasch, Diego Moreda, Fabio Carta, Julian B\u00fcchel, Corey Lammie, Charles Mackin, Kim Tran, Tayfun Gokmen, Manuel Le Gallo-Bourdeau, and Kaoutar El Maghraoui\nas the initial core authors, along with many [contributors].\n\nYou can contact us by opening a new issue in the repository or alternatively\nat the ``aihwkit@us.ibm.com`` email address.\n\n## License\n\nThis project is licensed under [Apache License 2.0].\n\n[Apache License 2.0]: LICENSE.txt\n[Python package index]: https://pypi.org/project/aihwkit\n[`PyTorch`]: https://pytorch.org/\n\n[`examples/`]: examples/\n[`notebooks/`]: notebooks/\n[documentation]: https://aihwkit.readthedocs.io/\n[contributors]: https://github.com/IBM/aihwkit/graphs/contributors\n[advanced installation]: https://aihwkit.readthedocs.io/en/latest/advanced_install.html\n\n[von Neumann bottleneck]: https://en.wikipedia.org/wiki/Von_Neumann_architecture#Von_Neumann_bottleneck\n[phase-change memory]: https://en.wikipedia.org/wiki/Phase-change_memory\n[resistive random-access memory]: https://en.wikipedia.org/wiki/Resistive_random-access_memory\n[Flash memory]: https://en.wikipedia.org/wiki/Flash_memory\n[Kirchhoff\u2019s circuits laws]: https://en.wikipedia.org/wiki/Kirchhoff%27s_circuit_laws\n[online demo]: https://analog-ai-demo.mybluemix.net/\n[AIHW Composer]: https://aihw-composer.draco.res.ibm.com\n[award]: https://conferences.computer.org/services/2023/awards/\n[CUDA Dockerfile instructions]: https://github.com/IBM/aihwkit/blob/master/docs/source/advanced_install.rst#cuda-enabled-docker-image\n",
"bugtrack_url": null,
"license": "Apache 2.0",
"summary": "IBM Analog Hardware Acceleration Kit",
"version": "0.9.2",
"project_urls": {
"Homepage": "https://github.com/IBM/aihwkit"
},
"split_keywords": [
"ai",
" analog",
" rpu",
" torch",
" memristor",
" pcm",
" reram",
" crossbar",
" in-memory",
" nvm",
" non-von-neumann",
" non-volatile memory",
" phase-change material"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "33c5680f19cca0b33e648e42627f4c168806c367fe9254bc2c97db2b024ddda8",
"md5": "fa6eb8a488124f0bdeac6f86b3b82ee3",
"sha256": "15b997086f7a62213bccdfe06c8506cf8588ad7aebc44a572a84a3e8e1ea7395"
},
"downloads": -1,
"filename": "aihwkit-0.9.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"has_sig": false,
"md5_digest": "fa6eb8a488124f0bdeac6f86b3b82ee3",
"packagetype": "bdist_wheel",
"python_version": "cp310",
"requires_python": ">=3.7",
"size": 11331088,
"upload_time": "2024-09-19T11:20:27",
"upload_time_iso_8601": "2024-09-19T11:20:27.013122Z",
"url": "https://files.pythonhosted.org/packages/33/c5/680f19cca0b33e648e42627f4c168806c367fe9254bc2c97db2b024ddda8/aihwkit-0.9.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "6600bfb1caa379227f2cab057c20405f50a451d665992084619d43d59a9aedaa",
"md5": "f1b722602296e0d450e5c181efb070ae",
"sha256": "8c1e2f4202afa307019dd82a4bf0d3877a57c933b38a36b3cbc83f37094e5fe5"
},
"downloads": -1,
"filename": "aihwkit-0.9.2-cp38-cp38-macosx_10_9_x86_64.whl",
"has_sig": false,
"md5_digest": "f1b722602296e0d450e5c181efb070ae",
"packagetype": "bdist_wheel",
"python_version": "cp38",
"requires_python": ">=3.7",
"size": 10726006,
"upload_time": "2024-09-19T11:20:31",
"upload_time_iso_8601": "2024-09-19T11:20:31.364339Z",
"url": "https://files.pythonhosted.org/packages/66/00/bfb1caa379227f2cab057c20405f50a451d665992084619d43d59a9aedaa/aihwkit-0.9.2-cp38-cp38-macosx_10_9_x86_64.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "38bffd6451c6cc50b5b8d005ec074af7e0eddf7a461fd1859cf3e64a51547156",
"md5": "d58f287c98f72c8d2b6114dd850b3b3d",
"sha256": "ed5fedb8f90961607e55f81609f85a8673190b5644d968b5cf068eb60639d6fe"
},
"downloads": -1,
"filename": "aihwkit-0.9.2-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"has_sig": false,
"md5_digest": "d58f287c98f72c8d2b6114dd850b3b3d",
"packagetype": "bdist_wheel",
"python_version": "cp38",
"requires_python": ">=3.7",
"size": 11330821,
"upload_time": "2024-09-19T11:20:35",
"upload_time_iso_8601": "2024-09-19T11:20:35.839066Z",
"url": "https://files.pythonhosted.org/packages/38/bf/fd6451c6cc50b5b8d005ec074af7e0eddf7a461fd1859cf3e64a51547156/aihwkit-0.9.2-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "744672a05502f14f4ddf7a841a645f528d6fe5acb509139680ee3d5facfab4f6",
"md5": "a364fd4c5fdac5a6c4f2e48dea8ae6f7",
"sha256": "2131af4f713a699dd4c58e9fdb720ac09b1649a78cf4fe33d73b9d59e21eb9af"
},
"downloads": -1,
"filename": "aihwkit-0.9.2-cp39-cp39-macosx_10_9_x86_64.whl",
"has_sig": false,
"md5_digest": "a364fd4c5fdac5a6c4f2e48dea8ae6f7",
"packagetype": "bdist_wheel",
"python_version": "cp39",
"requires_python": ">=3.7",
"size": 10726332,
"upload_time": "2024-09-19T11:20:39",
"upload_time_iso_8601": "2024-09-19T11:20:39.981689Z",
"url": "https://files.pythonhosted.org/packages/74/46/72a05502f14f4ddf7a841a645f528d6fe5acb509139680ee3d5facfab4f6/aihwkit-0.9.2-cp39-cp39-macosx_10_9_x86_64.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "bb644f93ca618590bc773ae5adf8511c950b1733781bdcb18c49ef5db24b5a0e",
"md5": "a565e7c154604c34579dfccd258a9069",
"sha256": "0e186a4758720144404c9975ca217ff631f08855f6ec80c9e70aa36a79ae611c"
},
"downloads": -1,
"filename": "aihwkit-0.9.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"has_sig": false,
"md5_digest": "a565e7c154604c34579dfccd258a9069",
"packagetype": "bdist_wheel",
"python_version": "cp39",
"requires_python": ">=3.7",
"size": 11331173,
"upload_time": "2024-09-19T11:20:44",
"upload_time_iso_8601": "2024-09-19T11:20:44.138588Z",
"url": "https://files.pythonhosted.org/packages/bb/64/4f93ca618590bc773ae5adf8511c950b1733781bdcb18c49ef5db24b5a0e/aihwkit-0.9.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "7bed385adb141f3691000c33adb3d76c80e1fe213d3e991b91054066e4f698b6",
"md5": "361b706d8585033586f0797dc4068040",
"sha256": "cbcc4410830786edb510564dcbdc4c289490b9099d43535ba71bb8de999fef26"
},
"downloads": -1,
"filename": "aihwkit-0.9.2.tar.gz",
"has_sig": false,
"md5_digest": "361b706d8585033586f0797dc4068040",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.7",
"size": 612504,
"upload_time": "2024-09-19T11:20:46",
"upload_time_iso_8601": "2024-09-19T11:20:46.635616Z",
"url": "https://files.pythonhosted.org/packages/7b/ed/385adb141f3691000c33adb3d76c80e1fe213d3e991b91054066e4f698b6/aihwkit-0.9.2.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-09-19 11:20:46",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "IBM",
"github_project": "aihwkit",
"travis_ci": true,
"coveralls": false,
"github_actions": false,
"requirements": [],
"lcname": "aihwkit"
}
JSON
