fastmtl

Name	fastmtl JSON
Version	1.1.0 JSON
	download
home_page	https://github.com/bdsaglam/fastmtl/tree/master/
Summary	Multi-task learning utilities for fastai
upload_time	2023-01-22 20:15:06
maintainer
docs_url	None
author	Barış Deniz Sağlam
requires_python	>=3.7
license	Apache Software License 2.0
keywords	deep learning multi-task learning fastai pytorch
VCS
bugtrack_url
requirements	No requirements were recorded.
Travis-CI	No Travis.
coveralls test coverage	No coveralls.

            # fastmtl
> Multi-task learning utilities for fastai


## Install

`pip install fastmtl`

## Usage

### Loss

Apply a loss function per model output and get weighted sum of them. For instance, if the first model output is for classification and the second model output is for regression,
```py
from fastmtl.loss import CombinedLoss
loss_func = CombinedLoss(CrossEntropyLossFlat(), MSELossFlat(), weight=[1.0, 3.0])
```

### Metric

Apply metrics for each model output. For instance, if we have a model making classification and regression, we can evaluate each model output with relevant metrics. Assuming that model outputs a tuple of tensors for classification and regression, respectively:

```py
from fastai.metrics import F1Score, R2Score
from fastmtl.metric import mtl_metrics

clf_f1_macro =  F1Score(average='macro')
clf_f1_macro.name = 'clf_f1(macro)'
clf_f1_micro =  F1Score(average='micro')
clf_f1_micro.name = 'clf_f1(micro)'

reg_r2 = R2Score()
reg_r2.name = 'reg_r2'

# metrics for classification in the first list 
# metrics for regression in the second list 
metrics = mtl_metrics([clf_f1_macro, clf_f1_micro], [reg_r2])

learn = Learner(
    ...
    metrics=metrics,
)
```

## Tutorials

[Video distortion detection](https://bdsaglam.github.io/fastmtl/tutorial.vqa)

## TODO
- [ ] Support tabular learner
- [ ] Support fastai>=2.7



            

Raw data

            {
    "_id": null,
    "home_page": "https://github.com/bdsaglam/fastmtl/tree/master/",
    "name": "fastmtl",
    "maintainer": "",
    "docs_url": null,
    "requires_python": ">=3.7",
    "maintainer_email": "",
    "keywords": "deep learning,multi-task learning,fastai,pytorch",
    "author": "Bar\u0131\u015f Deniz Sa\u011flam",
    "author_email": "bdsaglam@gmail.com",
    "download_url": "https://files.pythonhosted.org/packages/f3/86/8c7cbdef0c43e1e72af32f15ad22f83ac8e834627fec029c773f5d67ddd2/fastmtl-1.1.0.tar.gz",
    "platform": null,
    "description": "# fastmtl\n> Multi-task learning utilities for fastai\n\n\n## Install\n\n`pip install fastmtl`\n\n## Usage\n\n### Loss\n\nApply a loss function per model output and get weighted sum of them. For instance, if the first model output is for classification and the second model output is for regression,\n```py\nfrom fastmtl.loss import CombinedLoss\nloss_func = CombinedLoss(CrossEntropyLossFlat(), MSELossFlat(), weight=[1.0, 3.0])\n```\n\n### Metric\n\nApply metrics for each model output. For instance, if we have a model making classification and regression, we can evaluate each model output with relevant metrics. Assuming that model outputs a tuple of tensors for classification and regression, respectively:\n\n```py\nfrom fastai.metrics import F1Score, R2Score\nfrom fastmtl.metric import mtl_metrics\n\nclf_f1_macro =  F1Score(average='macro')\nclf_f1_macro.name = 'clf_f1(macro)'\nclf_f1_micro =  F1Score(average='micro')\nclf_f1_micro.name = 'clf_f1(micro)'\n\nreg_r2 = R2Score()\nreg_r2.name = 'reg_r2'\n\n# metrics for classification in the first list \n# metrics for regression in the second list \nmetrics = mtl_metrics([clf_f1_macro, clf_f1_micro], [reg_r2])\n\nlearn = Learner(\n    ...\n    metrics=metrics,\n)\n```\n\n## Tutorials\n\n[Video distortion detection](https://bdsaglam.github.io/fastmtl/tutorial.vqa)\n\n## TODO\n- [ ] Support tabular learner\n- [ ] Support fastai>=2.7\n\n\n",
    "bugtrack_url": null,
    "license": "Apache Software License 2.0",
    "summary": "Multi-task learning utilities for fastai",
    "version": "1.1.0",
    "split_keywords": [
        "deep learning",
        "multi-task learning",
        "fastai",
        "pytorch"
    ],
    "urls": [
        {
            "comment_text": "",
            "digests": {
                "blake2b_256": "f3868c7cbdef0c43e1e72af32f15ad22f83ac8e834627fec029c773f5d67ddd2",
                "md5": "26054e2728635c8334bd0a56226e5fab",
                "sha256": "a223f1c57c89bd173e2e68a2af1acc98b0977843c12f46b1c5ea66d560fd783b"
            },
            "downloads": -1,
            "filename": "fastmtl-1.1.0.tar.gz",
            "has_sig": false,
            "md5_digest": "26054e2728635c8334bd0a56226e5fab",
            "packagetype": "sdist",
            "python_version": "source",
            "requires_python": ">=3.7",
            "size": 10034,
            "upload_time": "2023-01-22T20:15:06",
            "upload_time_iso_8601": "2023-01-22T20:15:06.594228Z",
            "url": "https://files.pythonhosted.org/packages/f3/86/8c7cbdef0c43e1e72af32f15ad22f83ac8e834627fec029c773f5d67ddd2/fastmtl-1.1.0.tar.gz",
            "yanked": false,
            "yanked_reason": null
        }
    ],
    "upload_time": "2023-01-22 20:15:06",
    "github": false,
    "gitlab": false,
    "bitbucket": false,
    "lcname": "fastmtl"
}