This is a one stop solution for creating v2 requests, in the most basic format ever! Simply send in the deployed model urls and your inputs and watch the magic happen. Returns Json formatted output data with interpretable code which makes it even easier to use with almost every huggingface model.
Deployed Url:- https://v2.demo1.truefoundry.com/predict
https://huggingface.co/sshleifer/tiny-gpt2
text-generationhttps://huggingface.co/d4data/biomedical-ner-all/tree/main
token-classificationhttps://huggingface.co/typeform/distilbert-base-uncased-mnli
zero-shot-classificationhttps://huggingface.co/TahaDouaji/detr-doc-table-detection
object-detectionInput Format:
{
"hf_pipeline": string,
"model_deployed_url": string,
"inputs": any,
"parameters":any
}cUrls can also be used similarly
Object detection:
import requests
headers = {
'accept': 'application/json',
'Content-Type': 'application/json',
}
data = '{ "hf_pipeline": "object-detection","model_deployed_url": "https://object-intern-yajat.demo1.truefoundry.com/v2/models/object/infer", "inputs": "https://www.w3.org/WAI/WCAG22/Techniques/pdf/img/table-word.jpg","parameters": {}}'
response = requests.post('https://v2.demo1.truefoundry.com/predict', headers=headers, data=data)
print(response.json())Output:
[
{
"score":0.9906801581382751,
"label":"table",
"box":{
"xmin":20,
"ymin":54,
"xmax":479,
"ymax":257
}
}
]Text Generation:
import requests
headers = {
'accept': 'application/json',
'Content-Type': 'application/json',
}
data = '{"hf_pipeline": "text-generation","model_deployed_url":"https://text-generation-ml-intern-assign.tfy-gcp-standard-usce1.devtest.truefoundry.tech/v2/models/text-generation/infer","inputs": "Hello, how are you today? ","parameters": {"min_new_tokens": 10,"do_sample": true,"temperature": 1.0,"max_new_tokens": 20,"num_return_sequences": 5}}'
response = requests.post('https://v2.demo1.truefoundry.com/predict', headers=headers, data=data)
print(response.json())Output:
[
{
"generated_text":"Hello how are you today?PocketProsPros Late Boone� brutality skilletOutside omega Late workshops omega representations predators incarcer WheelsGy PocketGy Pocket653MostOutside representations factors Bend grandchildren praying boils Medicacious Redux mutual omega skillet Pocket predators predators� representations Tre"
}
]To get started on your own pc,
- Clone the repo:
git clone https://github.com/yxjat/simple-v2-wrapper- Run the uvicorn server
uvicorn main:app --host 0.0.0.0 --port 8000And you're all set!
https://object-intern-yajat.demo1.truefoundry.comhttps://textgen-intern-yajat.demo1.truefoundry.comNote! the object detection pipeline might take a while to respond (10s)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent