from huggingface_hub import InferenceClient, HfApi, upload_file
import datetime
import gradio as gr
import requests
import random
import prompts
import json
import uuid
import os
token=os.environ.get("HF_TOKEN")
username="omnibus"
dataset_name="tmp"
save_data=f'https://huggingface.co/datasets/{username}/{dataset_name}/raw/main/'
api=HfApi(token="")
VERBOSE=False
history = []
hist_out= []
summary =[]
main_point=[]
summary.append("")
main_point.append("")
list_of_users=["user1","user2","user3"]
persona=[
{"name":"Mr. Nice Guy", "description":"Nice","personality":"friendly, caring, helpful and informative. You always compliment people, and stick up for them, and you have no patience for bullies."},
{"name":"Mr. Mean Guy", "description":"Mean","personality":"a total asshole. You think you are really smart, but really you are just ignorant and mean. You don't have time for everybodies stupidity, and you let them know that in the comments."},
{"name":"Smarty Pants", "description":"Genius","personality":"intelligent, informative, know-it-all. You are the smartest guy in the room and always one-up the blog poster to show how mart you are."},
{"name":"Try Hard", "description":"Not Genius","personality":"dimwitted, lacking understanding about any topic. You always ask really irrelevant questions about the post."},
{"name":"Class Clown", "description":"Humerous","personality":"humerous, funny. You turn everything into a joke. Make a joke about the post."},
]
persona_names=[]
for ea in persona:
persona_names.append(ea['name'])
models=[
"mistralai/Mixtral-8x7B-Instruct-v0.1",
"mistralai/Mixtral-8x7B-Instruct-v0.2",
"google/gemma-7b",
"google/gemma-7b-it",
"google/gemma-2b",
"google/gemma-2b-it",
"meta-llama/Llama-2-7b-chat-hf",
"codellama/CodeLlama-70b-Instruct-hf",
"openchat/openchat-3.5-0106",
"NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
]
client_z=[]
def load_models(inp):
if VERBOSE==True:
print(type(inp))
print(inp)
print(models[inp])
client_z.clear()
client_z.append(InferenceClient(models[inp]))
#if "mistralai" in models[inp]:
# tokens = gr.Slider(label="Max new tokens",value=1600,minimum=0,maximum=8000,step=64,interactive=True, visible=True,info="The maximum number of tokens")
return gr.update(label=models[inp])
def format_prompt(message, history):
prompt = ""
for user_prompt, bot_response in history:
prompt += f"[INST] {user_prompt} [/INST]"
prompt += f" {bot_response} "
prompt += f"[INST] {message} [/INST]"
return prompt
agents =[
"COMMENTER",
"BLOG_POSTER",
"REPLY_TO_COMMENTER",
"COMPRESS_HISTORY_PROMPT"
]
temperature=0.9
max_new_tokens=256
max_new_tokens2=4000
top_p=0.95
repetition_penalty=1.0,
def compress_history(formatted_prompt):
print("###############\nRUNNING COMPRESS HISTORY\n###############\n")
seed = random.randint(1,1111111111111111)
agent=prompts.COMPRESS_HISTORY_PROMPT.format(history=summary[0],focus=main_point[0])
system_prompt=agent
temperature = 0.9
if temperature < 1e-2:
temperature = 1e-2
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=1048,
top_p=0.95,
repetition_penalty=1.0,
do_sample=True,
seed=seed,
)
#history.append((prompt,""))
#formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
formatted_prompt = formatted_prompt
client=client_z[0]
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
output += response.token.text
#history.append((output,history))
print(output)
print(main_point[0])
return output
def comment_generate(prompt, history,post_check,full_conv,persona2, agent_name=agents[0], sys_prompt="", temperature=0.9, max_new_tokens=1028, top_p=0.95, repetition_penalty=1.3,):
#def question_generate(prompt, history):
uid=uuid.uuid4()
print(post_check)
#full_conv=history
print(f'full_conv::\n{full_conv}')
print("###############\nRUNNING QUESTION GENERATOR\n###############\n")
seed = random.randint(1,1111111111111111)
#agent=prompts.COMMENTER.format(focus=main_point[0])
#bp,_,_=full_conv[0]
agent=prompts.COMMENTER.format(focus=full_conv[0][0],persona=persona[persona2]['personality'])
system_prompt=agent
temperature = float(temperature)
if temperature < 1e-2:
temperature = 1e-2
top_p = float(top_p)
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=max_new_tokens,
top_p=top_p,
repetition_penalty=repetition_penalty,
do_sample=True,
seed=seed,
)
#history.append((prompt,""))
#formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
#[INST] {message} [/INST]
formatted_prompt = f"[INST] {system_prompt}, {prompt} [/INST]"
client=client_z[0]
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
output += response.token.text
history.append((output,None))
comment_cnt=post_check['comment']
print(type(comment_cnt))
post_check['comment']=comment_cnt+1
#out_json = {'user':"",'datetime':current_time,'title':title,'blog':1,'comment':0,'reply':0,"prompt":prompt,"output":output}
#full_conv[-1]+=(output,)
full_conv.append((None,output,None))
html_out=load_html(full_conv,post_check['title'],post_check['user'])
#out_json = {'user':list_of_users[0],'datetime':current_time,'file_name':filename,'title':title,'blog':1,'comment':0,'reply':0,"prompt":prompt,"output":output,'comment_list':[]}
file_n = f'{post_check["filename"]}.json'
print(file_n)
r = requests.get(f'{save_data}book1/{file_n}')
print(f'status code main:: {r.status_code}')
if r.status_code==200:
try:
lod = json.loads(r.text)
print(f'lod:: {lod}')
lod[0]['comment']=lod[0]['comment']+1
lod[0]['comment_list'].append({'user':persona[persona2]['name'],'datetime':'','comment':output,'reply_list':[]})
#hist_out.append(out_json)
#try:
# for ea in
with open(f'{uid}.json', 'w') as f:
json_hist=json.dumps(lod, indent=4)
f.write(json_hist)
f.close()
upload_file(
path_or_fileobj =f"{uid}.json",
path_in_repo = f"book1/{file_n}",
repo_id =f"{username}/{dataset_name}",
repo_type = "dataset",
token=token,
)
except Exception as e:
print(e)
return "",history,lod[0],lod[0],lod[0],html_out
def reply_generate(prompt, history,post_check,full_conv,persona1, agent_name=agents[0], sys_prompt="", temperature=0.9, max_new_tokens=1028, top_p=0.95, repetition_penalty=1.0,):
#def question_generate(prompt, history):
uid=uuid.uuid4()
print(post_check)
#full_conv=history
print(f'full_conv::\n{full_conv}')
print("###############\nRUNNING REPLY GENERATOR\n###############\n")
seed = random.randint(1,1111111111111111)
agent=prompts.REPLY_TO_COMMENTER.format(focus=full_conv[0][0],comment=full_conv[1][1],persona=persona[persona1]['personality'])
system_prompt=agent
temperature = float(temperature)
if temperature < 1e-2:
temperature = 1e-2
top_p = float(top_p)
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=max_new_tokens,
top_p=top_p,
repetition_penalty=repetition_penalty,
do_sample=True,
seed=seed,
)
#history.append((prompt,""))
#formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
formatted_prompt = f"[INST] {system_prompt}, {prompt} [/INST]"
client=client_z[0]
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
output += response.token.text
history.append((output,None))
reply_cnt=post_check['reply']
post_check['reply']=reply_cnt+1
#out_json = {'user':"",'datetime':current_time,'title':title,'blog':1,'comment':0,'reply':0,"prompt":prompt,"output":output}
#full_conv[-1]+=(output,)
full_conv.append((None,None,output))
html_out=load_html(full_conv,post_check['title'],post_check['user'])
file_n = f'{post_check["filename"]}.json'
print(file_n)
r = requests.get(f'{save_data}book1/{file_n}')
print(f'status code main:: {r.status_code}')
if r.status_code==200:
try:
lod = json.loads(r.text)
print(f'lod:: {lod}')
lod[0]['reply']=lod[0]['reply']+1
lod[0]['comment_list'][0]['reply_list'].append({'user':persona[persona1]['name'],'datetime':'','reply':output})
#hist_out.append(out_json)
#try:
# for ea in
with open(f'{uid}.json', 'w') as f:
json_hist=json.dumps(lod, indent=4)
f.write(json_hist)
f.close()
upload_file(
path_or_fileobj =f"{uid}.json",
path_in_repo = f"book1/{file_n}",
repo_id =f"{username}/{dataset_name}",
repo_type = "dataset",
token=token,
)
except Exception as e:
print(e)
return "",history,lod[0],lod[0],lod[0],html_out
def reply_generate_OG(prompt, history, agent_name=agents[0], sys_prompt="", temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,):
#def question_generate(prompt, history):
print("###############\nRUNNING BLOG POSTER REPLY\n###############\n")
seed = random.randint(1,1111111111111111)
agent=prompts.REPLY_TO_COMMENTER.format(focus=main_point[0])
system_prompt=agent
temperature = float(temperature)
if temperature < 1e-2:
temperature = 1e-2
top_p = float(top_p)
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=max_new_tokens,
top_p=top_p,
repetition_penalty=repetition_penalty,
do_sample=True,
seed=seed,
)
#history.append((prompt,""))
formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
client=client_z[0]
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
output += response.token.text
#history.append((output,history))
return output
def create_valid_filename(invalid_filename: str) -> str:
"""Converts invalid characters in a string to be suitable for a filename."""
invalid_filename.replace(" ","-")
valid_chars = '-'.join(invalid_filename.split())
allowed_chars = ('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm',
'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z',
'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M',
'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z',
'0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '_', '-')
return ''.join(char for char in valid_chars if char in allowed_chars)
def load_html(inp,title,user_name="None"):
ht=""
if inp:
for i,ea in enumerate(inp):
blog,comm,repl=ea
#print(f'outp:: {outp}')
#print(f'prom:: {prom}')
ht+=f"""
{blog}""" if comm: ht+=f"""
{comm}""" if repl: ht+=f"""
{repl}""" ht+=f"""
{outp}
{prom}
{outp}
{prom}