A multimodal model training toolkit
Project description
Cornstarch
Build, Train, Run Your Own Multimodal Model
Cornstarch is a multimodal model training framework, including distributed training features with 5D parallelism (PP, TP, CP, DP, and modality parallelism). You can create your own multimodal model with a set of HuggingFace unimodal models and train it.
Cornstarch provides
- Pipeline Template and Heterogeneous Pipeline Parallelism: specify different pipeline templates and combine them to deploy heterogeneous pipeline parallel execution
- Composable multimodal model creation: specify your own multimodal models from a set of HuggingFace transformers unimodal models
- MultimodalModel Generation and Parallelization: specify your own multimodal model and parallelize it with 5D parallelism (DP+PP+TP+CP+and modality parallelism)
Install and Run
Please refer to our document!
Research
A technial report will be released very soon.
Contact
- Insu Jang (insujang@umich.edu)
Release history Release notifications | RSS feed
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 cornstarch-0.0.5.tar.gz.
File metadata
- Download URL: cornstarch-0.0.5.tar.gz
- Upload date:
- Size: 182.5 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.12.3
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
a93f485b384687bf8cf147f8ae23df0951241f30dfe46d00fac66f0912c0ca41
|
|
| MD5 |
d9bb0292e7587aa481fc0cc580593c5e
|
|
| BLAKE2b-256 |
19ee7fb0cdc927b1dd9bb4a7a0cdc7e9c121873383ff0f8ea318839d113e202e
|
File details
Details for the file cornstarch-0.0.5-py3-none-any.whl.
File metadata
- Download URL: cornstarch-0.0.5-py3-none-any.whl
- Upload date:
- Size: 270.3 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.12.3
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
ffbafc8bec0da8ba49bcd022c8d6a98a3078f6794c988703638d00d59ce401ac
|
|
| MD5 |
151126d4a9764a2f2a7b458537e44985
|
|
| BLAKE2b-256 |
c11fbb55b3d602fca2be3fa2ac5b952d246f4a141a0a35f538de294dcf32b286
|
