PygmalionAI

/

pygmalion-6b

like 736

Follow

Pygmalion 307

any help please ??

we found ColossalAI with EnergonAI that can achieve the API interface

is there anyway to run this model with it ?

what about try FastAPI?
i've set up chatbot api server using FastAPI.

can you help please with steps or the main idea how to do it ?

do you mean to build the API inside the python file before make the text generation using the model ?

i thought this way but i think it's too pare hands way so i asked if there any ready to use package or component

thank you for sharing your idea <3 appreciate it and i will give it a try

here is my code as a short version

~~~
device = 'cuda' if torch.cuda.is_available() else 'cpu'

app = FastAPI()

model_name = "PygmalionAI/pygmalion-6b"
gpt = transformers.AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = transformers.AutoTokenizer.from_pretrained(model_name)
gpt.to(device)



@app
	.post('/completion/')
async def chat(data:Data, request: Request):
    prompt = tokenizer(data.prompt, return_tensors='pt')
    prompt = {key: value.to(device) for key, value in prompt.items()}
    out = gpt.generate(**prompt, min_length=128, max_length=256, do_sample=True)
    completion = tokenizer.decode(out[0][len(prompt["input_ids"][0]):])
    return completion

then you post the request to your server like below

url = "http://your.ser.ver.ip:port/completion/"
res = requests.post(url, data=json.dumps(data))
print(res.text)

you must match the format between post data and api data

thank you very much <3

Please can you provide the full code for me and how i would implement this coming from a golang background. I appreciate the help