Mixtral-RPG / app.py
Omnibus's picture
Update app.py
e40b96e verified
raw
history blame
7.69 kB
from huggingface_hub import InferenceClient
import gradio as gr
import random
from prompts import GAME_MASTER, COMPRESS_HISTORY, ADJUST_STATS
def format_prompt(message, history):
prompt=""
prompt = "<s>"
for user_prompt, bot_response in history:
prompt += f"[INST] {user_prompt} [/INST]"
prompt += f" {bot_response}</s> "
prompt += f"[INST] {message} [/INST]"
return prompt
temperature=0.99
top_p=0.99
repetition_penalty=1.0
def compress_history(history,seed,temperature=temperature,top_p=top_p,repetition_penalty=repetition_penalty):
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
print("COMPRESSING")
formatted_prompt=f"{COMPRESS_HISTORY.format(history=history)}"
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=1024,
top_p=top_p,
repetition_penalty=repetition_penalty,
do_sample=True,
seed=seed,
)
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
return output
MAX_HISTORY=100
opts=[]
def generate(prompt, history,max_new_tokens,health,seed,temperature=temperature,top_p=top_p,repetition_penalty=repetition_penalty):
opts.clear()
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
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,
)
cnt=0
stats=health
history1=history
'''
stats="*******************\n"
for eac in health:
stats+=f'{eac}\n'
stats+="*******************\n"
'''
for ea in history:
print (ea)
for l in ea:
print (l)
cnt+=len(l.split("\n"))
print(f'cnt:: {cnt}')
if cnt > MAX_HISTORY:
history1 = compress_history(str(history), temperature, top_p, repetition_penalty)
formatted_prompt = format_prompt(f"{GAME_MASTER.format(history=history1,stats=stats,dice=random.randint(1,10))}, {prompt}", history)
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
if history:
yield [(prompt,output)],stats,None,None
else:
yield [(prompt,output)],stats,None,None
generate_kwargs2 = dict(
temperature=temperature,
max_new_tokens=128,
top_p=top_p,
repetition_penalty=repetition_penalty,
do_sample=True,
seed=random.randint(1,99999999999)
#seed=42,
)
#history=""
#formatted_prompt2 = format_prompt(f"{ADJUST_STATS.format(history=output,health=health)}, {prompt}", history)
#stream2 = client.text_generation(f"{ADJUST_STATS.format(history=output,health=health)}", **generate_kwargs2, stream=True, details=True, return_full_text=False)
#output2=""
#for response in stream2:
# output2 += response.token.text
lines = output.strip().strip("\n").split("\n")
skills=[]
skill_dict={}
option_drop=[]
new_stat="*******************\n"
for i,line in enumerate(lines):
if ":" in line:
for z in range(1,5):
try:
if f'{z}.' in lines[i+z]:
print(lines[i+z])
opts.append(lines[i+z])
except Exception as e:
print(e)
pass
if ": " in line[:20]:
try:
lab_1 = line.split(": ")[0]
skill_1 = line.split(": ")[1].split(" ")[0].split("<")[0]
skill_1=int(skill_1)
skill_dict[lab_1]=skill_1
#skill ={lab_1:skill_1}
new_stat += f'{lab_1}: {skill_1}\n'
print(skills)
except Exception as e:
print (f'--Error :: {e}')
print(f'Line:: {line}')
skills.append(skill_dict)
new_stat+="*******************\n"
stats=new_stat
option_drop=gr.Dropdown(label="Choices", choices=[e for e in opts])
if history:
history.append((prompt,output))
yield history,stats,skills,option_drop
else:
yield [(prompt,output)],stats,skills,option_drop
def clear_fn():
return None,None
base_stats=[
{},
]
text_stats='''*******************
Health: 100
Power: 20
Strength: 24
*******************
'''
rand_val=random.randint(1,99999999999)
def check_rand(inp,val):
if inp==True:
return gr.Slider(label="Seed", minimum=1, maximum=99999999999, value=random.randint(1,99999999999))
else:
return gr.Slider(label="Seed", minimum=1, maximum=99999999999, value=int(val))
with gr.Blocks() as app:
gr.HTML("""<center><h1>Mixtral 8x7B RPG</h1><h3>Role Playing Game Master</h3>""")
with gr.Group():
with gr.Row():
with gr.Column(scale=3):
chatbot = gr.Chatbot(label="Mixtral 8x7B Game Master",height=500, layout='panel', show_copy_button=True)
with gr.Row():
with gr.Column(scale=3):
opt=gr.Dropdown(label="Choices",choices=["Start a new game"],allow_custom_value=True, value="Start a new game", interactive=True)
#prompt=gr.Textbox(label = "Prompt", value="Start a new game")
with gr.Column(scale=2):
rand = gr.Checkbox(label="Random", value=False)
seed=gr.Slider(label="Seed", minimum=1, maximum=99999999999, value=rand_val)
#models_dd=gr.Dropdown(choices=[m for m in return_list],interactive=True)
with gr.Row():
button=gr.Button()
stop_button=gr.Button("Stop")
clear_btn = gr.Button("Clear")
with gr.Row():
tokens = gr.Slider(label="Max new tokens",value=2096,minimum=0,maximum=1048*10,step=64,interactive=False, visible=False,info="The maximum numbers of new tokens")
with gr.Column(scale=1):
char_stats=gr.Textbox(value=text_stats)
json_out=gr.JSON()
textboxes = []
if opts:
textboxes.clear()
for i in range(len(opts)-1):
t = gr.Button(f"{opts[i]}")
textboxes.append(t)
#text=gr.JSON()
#inp_query.change(search_models,inp_query,models_dd)
#test_b=test_btn.click(itt,url,e_box)
clear_btn.click(clear_fn,None,[opt,chatbot])
go=button.click(check_rand,[rand,seed],seed).then(generate,[opt,chatbot,tokens,char_stats,seed],[chatbot,char_stats,json_out,opt])
stop_button.click(None,None,None,cancels=[go])
app.launch(show_api=False)
'''
examples=[["Start the Game", None, None, None, None, None, ],
["Start a Game based in the year 1322", None, None, None, None, None,],
]
gr.ChatInterface(
fn=generate,
chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel"),
additional_inputs=additional_inputs,
title="Mixtral RPG Game Master",
examples=examples,
concurrency_limit=20,
).launch(share=True,show_api=True)
'''