<div style="display:flex; text-align:center;">
<img src="https://huggingface.co/front/assets/huggingface_logo.svg" width="100"/>
<img src="https://github.com/aws/sagemaker-inference-toolkit/raw/master/branding/icon/sagemaker-banner.png" width="450"/>
</div>
# SageMaker Hugging Face Inference Toolkit
[](https://pypi.python.org/pypi/sagemaker_huggingface_inference_toolkit) [](https://pypi.python.org/pypi/sagemaker_huggingface_inference_toolkit) [](https://github.com/python/black)
SageMaker Hugging Face Inference Toolkit is an open-source library for serving 🤗 Transformers and Diffusers models on Amazon SageMaker. This library provides default pre-processing, predict and postprocessing for certain 🤗 Transformers and Diffusers models and tasks. It utilizes the [SageMaker Inference Toolkit](https://github.com/aws/sagemaker-inference-toolkit) for starting up the model server, which is responsible for handling inference requests.
For Training, see [Run training on Amazon SageMaker](https://huggingface.co/docs/sagemaker/train).
For the Dockerfiles used for building SageMaker Hugging Face Containers, see [AWS Deep Learning Containers](https://github.com/aws/deep-learning-containers/tree/master/huggingface).
For information on running Hugging Face jobs on Amazon SageMaker, please refer to the [🤗 Transformers documentation](https://huggingface.co/docs/sagemaker).
For notebook examples: [SageMaker Notebook Examples](https://github.com/huggingface/notebooks/tree/master/sagemaker).
---
## 💻 Getting Started with 🤗 Inference Toolkit
_needs to be adjusted -> currently pseudo code_
**Install Amazon SageMaker Python SDK**
```bash
pip install sagemaker --upgrade
```
**Create a Amazon SageMaker endpoint with a trained model.**
```python
from sagemaker.huggingface import HuggingFaceModel
# create Hugging Face Model Class
huggingface_model = HuggingFaceModel(
transformers_version='4.6',
pytorch_version='1.7',
py_version='py36',
model_data='s3://my-trained-model/artifacts/model.tar.gz',
role=role,
)
# deploy model to SageMaker Inference
huggingface_model.deploy(initial_instance_count=1,instance_type="ml.m5.xlarge")
```
**Create a Amazon SageMaker endpoint with a model from the [🤗 Hub](https://huggingface.co/models).**
_note: This is an experimental feature, where the model will be loaded after the endpoint is created. Not all sagemaker features are supported, e.g. MME_
```python
from sagemaker.huggingface import HuggingFaceModel
# Hub Model configuration. https://huggingface.co/models
hub = {
'HF_MODEL_ID':'distilbert-base-uncased-distilled-squad',
'HF_TASK':'question-answering'
}
# create Hugging Face Model Class
huggingface_model = HuggingFaceModel(
transformers_version='4.6',
pytorch_version='1.7',
py_version='py36',
env=hub,
role=role,
)
# deploy model to SageMaker Inference
huggingface_model.deploy(initial_instance_count=1,instance_type="ml.m5.xlarge")
```
---
## 🛠️ Environment variables
The SageMaker Hugging Face Inference Toolkit implements various additional environment variables to simplify your deployment experience. A full list of environment variables is given below.
#### `HF_TASK`
The `HF_TASK` environment variable defines the task for the used 🤗 Transformers pipeline. A full list of tasks can be find [here](https://huggingface.co/transformers/main_classes/pipelines.html).
```bash
HF_TASK="question-answering"
```
#### `HF_MODEL_ID`
The `HF_MODEL_ID` environment variable defines the model id, which will be automatically loaded from [huggingface.co/models](https://huggingface.co/models) when creating or SageMaker Endpoint. The 🤗 Hub provides +10 000 models all available through this environment variable.
```bash
HF_MODEL_ID="distilbert-base-uncased-finetuned-sst-2-english"
```
#### `HF_MODEL_REVISION`
The `HF_MODEL_REVISION` is an extension to `HF_MODEL_ID` and allows you to define/pin a revision of the model to make sure you always load the same model on your SageMaker Endpoint.
```bash
HF_MODEL_REVISION="03b4d196c19d0a73c7e0322684e97db1ec397613"
```
#### `HF_API_TOKEN`
The `HF_API_TOKEN` environment variable defines the your Hugging Face authorization token. The `HF_API_TOKEN` is used as a HTTP bearer authorization for remote files, like private models. You can find your token at your [settings page](https://huggingface.co/settings/token).
```bash
HF_API_TOKEN="api_XXXXXXXXXXXXXXXXXXXXXXXXXXXXX"
```
#### `HF_TRUST_REMOTE_CODE`
The `HF_TRUST_REMOTE_CODE` environment variable defines wether or not to allow for custom models defined on the Hub in their own modeling files. Allowed values are `"True"` and `"False"`
```bash
HF_TRUST_REMOTE_CODE="True"
```
#### `HF_OPTIMUM_BATCH_SIZE`
The `HF_OPTIMUM_BATCH_SIZE` environment variable defines the batch size, which is used when compiling the model to Neuron. The default value is `1`. Not required when model is already converted.
```bash
HF_OPTIMUM_BATCH_SIZE="1"
```
#### `HF_OPTIMUM_SEQUENCE_LENGTH`
The `HF_OPTIMUM_SEQUENCE_LENGTH` environment variable defines the sequence length, which is used when compiling the model to Neuron. There is no default value. Not required when model is already converted.
```bash
HF_OPTIMUM_SEQUENCE_LENGTH="128"
```
---
## 🧑🏻💻 User defined code/modules
The Hugging Face Inference Toolkit allows user to override the default methods of the `HuggingFaceHandlerService`. Therefore, they need to create a folder named `code/` with an `inference.py` file in it. You can find an example for it in [sagemaker/17_customer_inference_script](https://github.com/huggingface/notebooks/blob/master/sagemaker/17_custom_inference_script/sagemaker-notebook.ipynb).
For example:
```bash
model.tar.gz/
|- pytorch_model.bin
|- ....
|- code/
|- inference.py
|- requirements.txt
```
In this example, `pytorch_model.bin` is the model file saved from training, `inference.py` is the custom inference module, and `requirements.txt` is a requirements file to add additional dependencies.
The custom module can override the following methods:
* `model_fn(model_dir)`: overrides the default method for loading the model, the return value `model` will be used in the `predict()` for predicitions. It receives argument the `model_dir`, the path to your unzipped `model.tar.gz`.
* `transform_fn(model, data, content_type, accept_type)`: overrides the default transform function with a custom implementation. Customers using this would have to implement `preprocess`, `predict` and `postprocess` steps in the `transform_fn`. **NOTE: This method can't be combined with `input_fn`, `predict_fn` or `output_fn` mentioned below.**
* `input_fn(input_data, content_type)`: overrides the default method for preprocessing, the return value `data` will be used in the `predict()` method for predicitions. The input is `input_data`, the raw body of your request and `content_type`, the content type form the request Header.
* `predict_fn(processed_data, model)`: overrides the default method for predictions, the return value `predictions` will be used in the `postprocess()` method. The input is `processed_data`, the result of the `preprocess()` method.
* `output_fn(prediction, accept)`: overrides the default method for postprocessing, the return value `result` will be the respond of your request(e.g.`JSON`). The inputs are `predictions`, the result of the `predict()` method and `accept` the return accept type from the HTTP Request, e.g. `application/json`
---
## 🤝 Contributing
Please read [CONTRIBUTING.md](https://github.com/aws/sagemaker-huggingface-inference-toolkit/blob/main/CONTRIBUTING.md)
for details on our code of conduct, and the process for submitting pull
requests to us.
---
## 📜 License
SageMaker Hugging Face Inference Toolkit is licensed under the Apache 2.0 License.
---
## 🧑🏻💻 Development Environment
Install all test and development packages with
```bash
pip3 install -e ".[test,dev]"
```
## Run Model Locally
1. manually change `MMS_CONFIG_FILE`
```
wget -O sagemaker-mms.properties https://raw.githubusercontent.com/aws/deep-learning-containers/master/huggingface/build_artifacts/inference/config.properties
```
2. Run Container, e.g. `text-to-image`
```
HF_MODEL_ID="stabilityai/stable-diffusion-xl-base-1.0" HF_TASK="text-to-image" python src/sagemaker_huggingface_inference_toolkit/serving.py
```
3. Adjust `handler_service.py` and comment out `if content_type in content_types.UTF8_TYPES:` thats needed for SageMaker but cannot be used locally
3. Send request
```
curl --request POST \
--url http://localhost:8080/invocations \
--header 'Accept: image/png' \
--header 'Content-Type: application/json' \
--data '"{\"inputs\": \"Camera\"}" \
--output image.png
```
Raw data
{
"_id": null,
"home_page": "https://github.com/aws/sagemaker-huggingface-inference-toolkit",
"name": "sagemaker-huggingface-inference-toolkit",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.6.0",
"maintainer_email": "",
"keywords": "NLP deep-learning transformer pytorch tensorflow BERT GPT GPT-2 AWS Amazon SageMaker Cloud",
"author": "HuggingFace and Amazon Web Services",
"author_email": "",
"download_url": "https://files.pythonhosted.org/packages/ae/5b/18710f8a40903ae613243e996b2927e513025a77280e107ed1e15181b38f/sagemaker-huggingface-inference-toolkit-2.3.0.tar.gz",
"platform": null,
"description": "<div style=\"display:flex; text-align:center;\">\n<img src=\"https://huggingface.co/front/assets/huggingface_logo.svg\" width=\"100\"/> \n<img src=\"https://github.com/aws/sagemaker-inference-toolkit/raw/master/branding/icon/sagemaker-banner.png\" width=\"450\"/>\n</div>\n\n\n\n\n# SageMaker Hugging Face Inference Toolkit \n\n[](https://pypi.python.org/pypi/sagemaker_huggingface_inference_toolkit) [](https://pypi.python.org/pypi/sagemaker_huggingface_inference_toolkit) [](https://github.com/python/black)\n\n\nSageMaker Hugging Face Inference Toolkit is an open-source library for serving \ud83e\udd17 Transformers and Diffusers models on Amazon SageMaker. This library provides default pre-processing, predict and postprocessing for certain \ud83e\udd17 Transformers and Diffusers models and tasks. It utilizes the [SageMaker Inference Toolkit](https://github.com/aws/sagemaker-inference-toolkit) for starting up the model server, which is responsible for handling inference requests.\n\nFor Training, see [Run training on Amazon SageMaker](https://huggingface.co/docs/sagemaker/train).\n\nFor the Dockerfiles used for building SageMaker Hugging Face Containers, see [AWS Deep Learning Containers](https://github.com/aws/deep-learning-containers/tree/master/huggingface).\n\nFor information on running Hugging Face jobs on Amazon SageMaker, please refer to the [\ud83e\udd17 Transformers documentation](https://huggingface.co/docs/sagemaker).\n\nFor notebook examples: [SageMaker Notebook Examples](https://github.com/huggingface/notebooks/tree/master/sagemaker).\n\n---\n## \ud83d\udcbb Getting Started with \ud83e\udd17 Inference Toolkit\n\n_needs to be adjusted -> currently pseudo code_\n\n**Install Amazon SageMaker Python SDK**\n\n```bash\npip install sagemaker --upgrade\n```\n\n**Create a Amazon SageMaker endpoint with a trained model.**\n\n```python\nfrom sagemaker.huggingface import HuggingFaceModel\n\n# create Hugging Face Model Class\nhuggingface_model = HuggingFaceModel(\n transformers_version='4.6',\n pytorch_version='1.7',\n py_version='py36',\n model_data='s3://my-trained-model/artifacts/model.tar.gz',\n role=role,\n)\n# deploy model to SageMaker Inference\nhuggingface_model.deploy(initial_instance_count=1,instance_type=\"ml.m5.xlarge\")\n```\n\n\n**Create a Amazon SageMaker endpoint with a model from the [\ud83e\udd17 Hub](https://huggingface.co/models).** \n_note: This is an experimental feature, where the model will be loaded after the endpoint is created. Not all sagemaker features are supported, e.g. MME_\n```python\nfrom sagemaker.huggingface import HuggingFaceModel\n# Hub Model configuration. https://huggingface.co/models\nhub = {\n 'HF_MODEL_ID':'distilbert-base-uncased-distilled-squad',\n 'HF_TASK':'question-answering'\n}\n# create Hugging Face Model Class\nhuggingface_model = HuggingFaceModel(\n transformers_version='4.6',\n pytorch_version='1.7',\n py_version='py36',\n env=hub,\n role=role,\n)\n# deploy model to SageMaker Inference\nhuggingface_model.deploy(initial_instance_count=1,instance_type=\"ml.m5.xlarge\")\n```\n\n---\n\n## \ud83d\udee0\ufe0f Environment variables\n\nThe SageMaker Hugging Face Inference Toolkit implements various additional environment variables to simplify your deployment experience. A full list of environment variables is given below.\n\n#### `HF_TASK`\n\nThe `HF_TASK` environment variable defines the task for the used \ud83e\udd17 Transformers pipeline. A full list of tasks can be find [here](https://huggingface.co/transformers/main_classes/pipelines.html).\n\n```bash\nHF_TASK=\"question-answering\"\n```\n\n#### `HF_MODEL_ID`\n\nThe `HF_MODEL_ID` environment variable defines the model id, which will be automatically loaded from [huggingface.co/models](https://huggingface.co/models) when creating or SageMaker Endpoint. The \ud83e\udd17 Hub provides +10 000 models all available through this environment variable.\n\n```bash\nHF_MODEL_ID=\"distilbert-base-uncased-finetuned-sst-2-english\"\n```\n\n#### `HF_MODEL_REVISION`\n\nThe `HF_MODEL_REVISION` is an extension to `HF_MODEL_ID` and allows you to define/pin a revision of the model to make sure you always load the same model on your SageMaker Endpoint.\n\n```bash\nHF_MODEL_REVISION=\"03b4d196c19d0a73c7e0322684e97db1ec397613\"\n```\n\n#### `HF_API_TOKEN`\n\nThe `HF_API_TOKEN` environment variable defines the your Hugging Face authorization token. The `HF_API_TOKEN` is used as a HTTP bearer authorization for remote files, like private models. You can find your token at your [settings page](https://huggingface.co/settings/token).\n\n```bash\nHF_API_TOKEN=\"api_XXXXXXXXXXXXXXXXXXXXXXXXXXXXX\"\n```\n\n#### `HF_TRUST_REMOTE_CODE`\n\nThe `HF_TRUST_REMOTE_CODE` environment variable defines wether or not to allow for custom models defined on the Hub in their own modeling files. Allowed values are `\"True\"` and `\"False\"`\n\n```bash\nHF_TRUST_REMOTE_CODE=\"True\"\n```\n\n#### `HF_OPTIMUM_BATCH_SIZE`\n\nThe `HF_OPTIMUM_BATCH_SIZE` environment variable defines the batch size, which is used when compiling the model to Neuron. The default value is `1`. Not required when model is already converted. \n\n```bash\nHF_OPTIMUM_BATCH_SIZE=\"1\"\n```\n\n#### `HF_OPTIMUM_SEQUENCE_LENGTH`\n\nThe `HF_OPTIMUM_SEQUENCE_LENGTH` environment variable defines the sequence length, which is used when compiling the model to Neuron. There is no default value. Not required when model is already converted. \n\n```bash\nHF_OPTIMUM_SEQUENCE_LENGTH=\"128\"\n```\n\n---\n\n## \ud83e\uddd1\ud83c\udffb\u200d\ud83d\udcbb User defined code/modules\n\nThe Hugging Face Inference Toolkit allows user to override the default methods of the `HuggingFaceHandlerService`. Therefore, they need to create a folder named `code/` with an `inference.py` file in it. You can find an example for it in [sagemaker/17_customer_inference_script](https://github.com/huggingface/notebooks/blob/master/sagemaker/17_custom_inference_script/sagemaker-notebook.ipynb).\nFor example: \n```bash\nmodel.tar.gz/\n|- pytorch_model.bin\n|- ....\n|- code/\n |- inference.py\n |- requirements.txt \n```\nIn this example, `pytorch_model.bin` is the model file saved from training, `inference.py` is the custom inference module, and `requirements.txt` is a requirements file to add additional dependencies.\nThe custom module can override the following methods: \n\n* `model_fn(model_dir)`: overrides the default method for loading the model, the return value `model` will be used in the `predict()` for predicitions. It receives argument the `model_dir`, the path to your unzipped `model.tar.gz`.\n* `transform_fn(model, data, content_type, accept_type)`: overrides the default transform function with a custom implementation. Customers using this would have to implement `preprocess`, `predict` and `postprocess` steps in the `transform_fn`. **NOTE: This method can't be combined with `input_fn`, `predict_fn` or `output_fn` mentioned below.** \n* `input_fn(input_data, content_type)`: overrides the default method for preprocessing, the return value `data` will be used in the `predict()` method for predicitions. The input is `input_data`, the raw body of your request and `content_type`, the content type form the request Header.\n* `predict_fn(processed_data, model)`: overrides the default method for predictions, the return value `predictions` will be used in the `postprocess()` method. The input is `processed_data`, the result of the `preprocess()` method.\n* `output_fn(prediction, accept)`: overrides the default method for postprocessing, the return value `result` will be the respond of your request(e.g.`JSON`). The inputs are `predictions`, the result of the `predict()` method and `accept` the return accept type from the HTTP Request, e.g. `application/json`\n\n\n\n\n---\n## \ud83e\udd1d Contributing\n\nPlease read [CONTRIBUTING.md](https://github.com/aws/sagemaker-huggingface-inference-toolkit/blob/main/CONTRIBUTING.md)\nfor details on our code of conduct, and the process for submitting pull\nrequests to us.\n\n---\n## \ud83d\udcdc License\n\nSageMaker Hugging Face Inference Toolkit is licensed under the Apache 2.0 License.\n\n---\n\n## \ud83e\uddd1\ud83c\udffb\u200d\ud83d\udcbb Development Environment\n\nInstall all test and development packages with \n\n```bash\npip3 install -e \".[test,dev]\"\n```\n## Run Model Locally\n\n1. manually change `MMS_CONFIG_FILE`\n```\nwget -O sagemaker-mms.properties https://raw.githubusercontent.com/aws/deep-learning-containers/master/huggingface/build_artifacts/inference/config.properties\n```\n\n2. Run Container, e.g. `text-to-image`\n```\nHF_MODEL_ID=\"stabilityai/stable-diffusion-xl-base-1.0\" HF_TASK=\"text-to-image\" python src/sagemaker_huggingface_inference_toolkit/serving.py\n```\n3. Adjust `handler_service.py` and comment out `if content_type in content_types.UTF8_TYPES:` thats needed for SageMaker but cannot be used locally\n\n3. Send request \n```\ncurl --request POST \\\n --url http://localhost:8080/invocations \\\n --header 'Accept: image/png' \\\n --header 'Content-Type: application/json' \\\n --data '\"{\\\"inputs\\\": \\\"Camera\\\"}\" \\\n --output image.png\n```\n",
"bugtrack_url": null,
"license": "Apache License 2.0",
"summary": "Open source library for running inference workload with Hugging Face Deep Learning Containers on Amazon SageMaker.",
"version": "2.3.0",
"project_urls": {
"Homepage": "https://github.com/aws/sagemaker-huggingface-inference-toolkit"
},
"split_keywords": [
"nlp",
"deep-learning",
"transformer",
"pytorch",
"tensorflow",
"bert",
"gpt",
"gpt-2",
"aws",
"amazon",
"sagemaker",
"cloud"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "492d3f4bae8b62aefac3f2640ecba731d85d4f53f186b52900ee628f0e094c6f",
"md5": "dd8db5b8b63470e55aa076788e05ca0c",
"sha256": "1c4faf3e978914090a57d814d0be53b357a3cbc1c886c7f8bf6ac18c22522f17"
},
"downloads": -1,
"filename": "sagemaker_huggingface_inference_toolkit-2.3.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "dd8db5b8b63470e55aa076788e05ca0c",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.6.0",
"size": 28306,
"upload_time": "2023-11-17T08:42:09",
"upload_time_iso_8601": "2023-11-17T08:42:09.803529Z",
"url": "https://files.pythonhosted.org/packages/49/2d/3f4bae8b62aefac3f2640ecba731d85d4f53f186b52900ee628f0e094c6f/sagemaker_huggingface_inference_toolkit-2.3.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "ae5b18710f8a40903ae613243e996b2927e513025a77280e107ed1e15181b38f",
"md5": "c3392ffdddd7b21abb5e35132b5687d9",
"sha256": "0a3264e63b6b27e269e69d289f7daa6d919b18cd8c79cf0022b9a7cec38b628e"
},
"downloads": -1,
"filename": "sagemaker-huggingface-inference-toolkit-2.3.0.tar.gz",
"has_sig": false,
"md5_digest": "c3392ffdddd7b21abb5e35132b5687d9",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.6.0",
"size": 25795,
"upload_time": "2023-11-17T08:42:11",
"upload_time_iso_8601": "2023-11-17T08:42:11.520697Z",
"url": "https://files.pythonhosted.org/packages/ae/5b/18710f8a40903ae613243e996b2927e513025a77280e107ed1e15181b38f/sagemaker-huggingface-inference-toolkit-2.3.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-11-17 08:42:11",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "aws",
"github_project": "sagemaker-huggingface-inference-toolkit",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "sagemaker-huggingface-inference-toolkit"
}
JSON
