text-generation-webui/api-example-stream.py

93 lines
2.7 KiB
Python
Raw Normal View History

2023-03-07 00:34:36 +01:00
'''
Contributed by SagsMug. Thank you SagsMug.
https://github.com/oobabooga/text-generation-webui/pull/175
'''
2023-03-07 04:23:36 +01:00
import asyncio
import json
2023-03-06 12:13:50 +01:00
import random
2023-03-07 04:23:36 +01:00
import string
2023-03-06 12:13:50 +01:00
import websockets
2023-03-07 04:23:36 +01:00
2023-04-16 17:12:31 +02:00
# Gradio changes this index from time to time. To rediscover it, set VISIBLE = False in
# modules/api.py and use the dev tools to inspect the request made after clicking on the
# button called "Run" at the bottom of the UI
GRADIO_FN = 34
2023-03-06 12:13:50 +01:00
def random_hash():
2023-03-06 23:52:26 +01:00
letters = string.ascii_lowercase + string.digits
return ''.join(random.choice(letters) for i in range(9))
2023-03-06 12:13:50 +01:00
2023-03-06 12:13:50 +01:00
async def run(context):
2023-03-06 23:52:26 +01:00
server = "127.0.0.1"
params = {
'max_new_tokens': 200,
'do_sample': True,
2023-04-20 00:49:52 +02:00
'temperature': 0.72,
'top_p': 0.73,
2023-03-06 23:52:26 +01:00
'typical_p': 1,
2023-04-20 00:49:52 +02:00
'repetition_penalty': 1.1,
'encoder_repetition_penalty': 1.0,
2023-03-06 23:52:26 +01:00
'top_k': 0,
'min_length': 0,
'no_repeat_ngram_size': 0,
'num_beams': 1,
'penalty_alpha': 0,
'length_penalty': 1,
'early_stopping': False,
2023-03-22 19:40:20 +01:00
'seed': -1,
'add_bos_token': True,
'truncation_length': 2048,
'ban_eos_token': False,
'skip_special_tokens': True,
2023-04-20 00:49:52 +02:00
'stopping_strings': [],
2023-03-06 23:52:26 +01:00
}
payload = json.dumps([context, params])
2023-03-06 23:52:26 +01:00
session = random_hash()
2023-03-06 12:13:50 +01:00
2023-03-06 23:52:26 +01:00
async with websockets.connect(f"ws://{server}:7860/queue/join") as websocket:
while content := json.loads(await websocket.recv()):
# Python3.10 syntax, replace with if elif on older
2023-03-06 23:52:26 +01:00
match content["msg"]:
case "send_hash":
await websocket.send(json.dumps({
"session_hash": session,
"fn_index": GRADIO_FN
2023-03-06 23:52:26 +01:00
}))
case "estimation":
pass
case "send_data":
await websocket.send(json.dumps({
"session_hash": session,
"fn_index": GRADIO_FN,
2023-03-06 23:52:26 +01:00
"data": [
payload
2023-03-06 23:52:26 +01:00
]
}))
case "process_starts":
pass
case "process_generating" | "process_completed":
yield content["output"]["data"][0]
# You can search for your desired end indicator and
2023-03-06 23:52:26 +01:00
# stop generation by closing the websocket here
if (content["msg"] == "process_completed"):
break
2023-03-06 12:13:50 +01:00
prompt = "What I would like to say is the following: "
2023-03-06 12:13:50 +01:00
async def get_result():
2023-03-06 23:52:26 +01:00
async for response in run(prompt):
# Print intermediate steps
print(response)
2023-03-06 12:13:50 +01:00
2023-03-06 23:52:26 +01:00
# Print final result
print(response)
2023-03-06 12:13:50 +01:00
2023-03-06 23:52:26 +01:00
asyncio.run(get_result())