| Name | RiskBERT JSON |
| Version |
0.0.9
JSON |
| download |
| home_page | |
| Summary | Generalized semantic regression with a BERT base. |
| upload_time | 2024-02-20 08:16:37 |
| maintainer | |
| docs_url | None |
| author | |
| requires_python | >=3.7 |
| license | BSD-3-Clause |
| keywords |
llm
bert
glm
|
| VCS |
|
| bugtrack_url |
|
| requirements |
No requirements were recorded.
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
# generalized-semantic-regression
RiskBERT is a significant step forward, making it easier than ever to incorporate text fragments into various applications, such as insurance frequency and severity models, or other GLM-based models. Feel free to explore and utilize RiskBERT for your text analysis needs.
To learn more about the RiskBERT implementation read this article: https://www.thebigdatablog.com/generalized-semantic-regression-using-contextual-embeddings/
Example:
`pip install RiskBERT`
```
from transformers import AutoTokenizer
import torch
from RiskBERT import glmModel, RiskBertModel
from RiskBERT import trainer, evaluate_model
from RiskBERT.simulation.data_functions import Data
from RiskBERT.utils import DataConstructor
# Set device to gpu if available
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# Init the model
model_dataset = Data(20000, scores=torch.tensor([[0.2],[0.4]]), weigth=5)
pre_model= "distilbert-base-uncased"
model = RiskBertModel(model=pre_model, input_dim=2, dropout=0.4, freeze_bert=True, mode="CLS")
tokenizer = AutoTokenizer.from_pretrained(pre_model)
# Train the model
model, Total_Loss, Validation_Loss, Test_Loss = trainer(model =model,
model_dataset=model_dataset,
epochs=100,
batch_size=1000,
evaluate_fkt=evaluate_model,
tokenizer=tokenizer,
optimizer=torch.optim.SGD(model.parameters(), lr=0.001),
device = device
)
# Predict from the model
my_data = DataConstructor(
sentences=[["Dies ist ein Test"],["Hallo Welt", "RiskBERT ist das Beste"]],
covariates=[[1,5],[2,6]],
tokenizer= tokenizer).prepare_for_model()
my_prediction=model(**my_data)
```
# Upload to pip
```
python -m pip install build twine
python -m build
twine check dist/*
twine upload dist/*`
````
Raw data
{
"_id": null,
"home_page": "",
"name": "RiskBERT",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.7",
"maintainer_email": "",
"keywords": "LLM,BERT,GLM",
"author": "",
"author_email": "Heiko Wagner <heikowagner@t-online.de>",
"download_url": "https://files.pythonhosted.org/packages/15/c0/c92d553c20a8e3878ccc44d9105839a20b24b7fcef1a8bf335c134908874/RiskBERT-0.0.9.tar.gz",
"platform": null,
"description": "# generalized-semantic-regression\nRiskBERT is a significant step forward, making it easier than ever to incorporate text fragments into various applications, such as insurance frequency and severity models, or other GLM-based models. Feel free to explore and utilize RiskBERT for your text analysis needs.\n\nTo learn more about the RiskBERT implementation read this article: https://www.thebigdatablog.com/generalized-semantic-regression-using-contextual-embeddings/\n\nExample: \n`pip install RiskBERT`\n\n```\nfrom transformers import AutoTokenizer\nimport torch\nfrom RiskBERT import glmModel, RiskBertModel\nfrom RiskBERT import trainer, evaluate_model\nfrom RiskBERT.simulation.data_functions import Data\nfrom RiskBERT.utils import DataConstructor\n\n# Set device to gpu if available\ndevice = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n\n# Init the model\nmodel_dataset = Data(20000, scores=torch.tensor([[0.2],[0.4]]), weigth=5)\npre_model= \"distilbert-base-uncased\"\nmodel = RiskBertModel(model=pre_model, input_dim=2, dropout=0.4, freeze_bert=True, mode=\"CLS\")\ntokenizer = AutoTokenizer.from_pretrained(pre_model)\n# Train the model\nmodel, Total_Loss, Validation_Loss, Test_Loss = trainer(model =model, \n model_dataset=model_dataset, \n epochs=100,\n batch_size=1000,\n evaluate_fkt=evaluate_model,\n tokenizer=tokenizer, \n optimizer=torch.optim.SGD(model.parameters(), lr=0.001),\n device = device\n )\n\n# Predict from the model\nmy_data = DataConstructor(\n sentences=[[\"Dies ist ein Test\"],[\"Hallo Welt\", \"RiskBERT ist das Beste\"]], \n covariates=[[1,5],[2,6]],\n tokenizer= tokenizer).prepare_for_model()\nmy_prediction=model(**my_data)\n\n```\n\n# Upload to pip\n```\npython -m pip install build twine\npython -m build\ntwine check dist/*\ntwine upload dist/*`\n````\n\n",
"bugtrack_url": null,
"license": "BSD-3-Clause",
"summary": "Generalized semantic regression with a BERT base.",
"version": "0.0.9",
"project_urls": null,
"split_keywords": [
"llm",
"bert",
"glm"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "35a0d5768ccba22360beca65a5b13a7a7f78da337ce74f334cef2e019ea3b837",
"md5": "50a9dd37387190fa30466f953d5585b0",
"sha256": "720f02e9303eb36911c36d73916111b29858aed4f1e0298902480b09710a8c3c"
},
"downloads": -1,
"filename": "RiskBERT-0.0.9-py3-none-any.whl",
"has_sig": false,
"md5_digest": "50a9dd37387190fa30466f953d5585b0",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.7",
"size": 16201,
"upload_time": "2024-02-20T08:16:35",
"upload_time_iso_8601": "2024-02-20T08:16:35.322771Z",
"url": "https://files.pythonhosted.org/packages/35/a0/d5768ccba22360beca65a5b13a7a7f78da337ce74f334cef2e019ea3b837/RiskBERT-0.0.9-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "15c0c92d553c20a8e3878ccc44d9105839a20b24b7fcef1a8bf335c134908874",
"md5": "a83a0fbc87dbcb32e4d0efde1d4b6975",
"sha256": "60497ca774844542f2984c3b1e1a47c22e563ceac72f596aac2899ff5e0ede08"
},
"downloads": -1,
"filename": "RiskBERT-0.0.9.tar.gz",
"has_sig": false,
"md5_digest": "a83a0fbc87dbcb32e4d0efde1d4b6975",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.7",
"size": 21637,
"upload_time": "2024-02-20T08:16:37",
"upload_time_iso_8601": "2024-02-20T08:16:37.018314Z",
"url": "https://files.pythonhosted.org/packages/15/c0/c92d553c20a8e3878ccc44d9105839a20b24b7fcef1a8bf335c134908874/RiskBERT-0.0.9.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-02-20 08:16:37",
"github": false,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"lcname": "riskbert"
}
