Spaces:
Runtime error
Runtime error
File size: 8,676 Bytes
e4e0162 ae1f197 c47b3e3 f29cff9 ae1f197 efb5973 838b0a0 ae1f197 f29cff9 c47b3e3 f29cff9 ae1f197 e4e0162 cd41564 e4e0162 0c404dc cd41564 0c404dc e4e0162 0c404dc e4e0162 0c404dc 125deff 0c404dc e4e0162 0c404dc e4e0162 0c404dc f29cff9 e4e0162 0c404dc cd41564 0c404dc e4e0162 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 |
import gradio as gr
from gradio_client import Client
from huggingface_hub import InferenceClient
import random
#ss_client = Client("https://omnibus-html-image-current-tab.hf.space/")
models=[
"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",
"mistralai/Mixtral-8x7B-Instruct-v0.1",
"mistralai/Mixtral-8x7B-Instruct-v0.2"
]
'''clients=[
InferenceClient(models[0]),
InferenceClient(models[1]),
InferenceClient(models[2]),
InferenceClient(models[3]),
]'''
client_z=[]
def load_models(inp):
out_box=[gr.Chatbot(),gr.Chatbot(),gr.Chatbot(),gr.Chatbot()]
print(type(inp))
print(inp)
print(models[inp[0]])
client_z.clear()
for z,ea in enumerate(inp):
client_z.append(InferenceClient(models[inp[z]]))
out_box[z]=(gr.update(label=models[inp[z]]))
return out_box[0],out_box[1],out_box[2],out_box[3]
def format_prompt(message, history):
prompt = ""
if history:
#<start_of_turn>userHow does the brain work?<end_of_turn><start_of_turn>model
for user_prompt, bot_response in history:
prompt += f"{user_prompt}\n"
print(prompt)
prompt += f"{bot_response}\n"
print(prompt)
prompt += f"<start_of_turn>user{message}<end_of_turn><start_of_turn>model"
print(prompt)
return prompt
mega_hist=[]
def chat_inf(system_prompt,prompt,history,client_choice,seed,temp,tokens,top_p,rep_p,hid_val):
if len(client_choice)>=hid_val:
#token max=8192
client=client_z[int(hid_val)-1]
if not mega_hist[hid_val-1]:
mega_hist.append([])
#history = []
hist_len=0
if mega_hist[hid_val-1]:
hist_len=len(mega_hist[hid_val-1])
print(hist_len)
in_len=len(system_prompt+prompt)+hist_len
print("\n#########"+str(in_len))
if (in_len+tokens) > 8000:
yield [(prompt,"Wait. I need to compress our Chat history...")]
#history=compress_history(history,client_choice,seed,temp,tokens,top_p,rep_p)
yield [(prompt,"History has been compressed, processing request...")]
generate_kwargs = dict(
temperature=temp,
max_new_tokens=tokens,
top_p=top_p,
repetition_penalty=rep_p,
do_sample=True,
seed=seed,
)
#formatted_prompt=prompt
formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", mega_hist[hid_val-1])
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
yield [(prompt,output)]
mega_hist[hid_val-1].append((prompt,output))
yield mega_hist[hid_val-1]
else:
yield None
def chat_inf_og(system_prompt,prompt,history,client_choice,seed,temp,tokens,top_p,rep_p,hid_val):
if len(client_choice)>=hid_val:
#token max=8192
client=client_z[int(hid_val)-1]
if not history:
history = []
hist_len=0
if history:
hist_len=len(history)
print(hist_len)
in_len=len(system_prompt+prompt)+hist_len
print("\n#########"+str(in_len))
if (in_len+tokens) > 8000:
yield [(prompt,"Wait. I need to compress our Chat history...")]
#history=compress_history(history,client_choice,seed,temp,tokens,top_p,rep_p)
yield [(prompt,"History has been compressed, processing request...")]
generate_kwargs = dict(
temperature=temp,
max_new_tokens=tokens,
top_p=top_p,
repetition_penalty=rep_p,
do_sample=True,
seed=seed,
)
#formatted_prompt=prompt
formatted_prompt = format_prompt(f"{system_prompt}, {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
yield [(prompt,output)]
history.append((prompt,output))
yield history
else:
yield None
def clear_fn():
return None,None,None
rand_val=random.randint(1,1111111111111111)
def check_rand(inp,val):
if inp==True:
return gr.Slider(label="Seed", minimum=1, maximum=1111111111111111, value=random.randint(1,1111111111111111))
else:
return gr.Slider(label="Seed", minimum=1, maximum=1111111111111111, value=int(val))
with gr.Blocks() as app:
gr.HTML("""<center><h1 style='font-size:xx-large;'>Google Gemma Models</h1><br><h3>running on Huggingface Inference Client</h3><br><h7>EXPERIMENTAL""")
with gr.Row():
chat_a = gr.Chatbot(height=500)
chat_b = gr.Chatbot(height=500)
with gr.Row():
chat_c = gr.Chatbot(height=500)
chat_d = gr.Chatbot(height=500)
with gr.Group():
with gr.Row():
with gr.Column(scale=3):
inp = gr.Textbox(label="Prompt")
sys_inp = gr.Textbox(label="System Prompt (optional)")
with gr.Row():
with gr.Column(scale=2):
btn = gr.Button("Chat")
with gr.Column(scale=1):
with gr.Group():
stop_btn=gr.Button("Stop")
clear_btn=gr.Button("Clear")
client_choice=gr.Dropdown(label="Models",type='index',choices=[c for c in models],max_choices=4,multiselect=True,interactive=True)
with gr.Column(scale=1):
with gr.Group():
rand = gr.Checkbox(label="Random Seed", value=True)
seed=gr.Slider(label="Seed", minimum=1, maximum=1111111111111111,step=1, value=rand_val)
tokens = gr.Slider(label="Max new tokens",value=3840,minimum=0,maximum=8000,step=64,interactive=True, visible=True,info="The maximum number of tokens")
temp=gr.Slider(label="Temperature",step=0.01, minimum=0.01, maximum=1.0, value=0.9)
top_p=gr.Slider(label="Top-P",step=0.01, minimum=0.01, maximum=1.0, value=0.9)
rep_p=gr.Slider(label="Repetition Penalty",step=0.1, minimum=0.1, maximum=2.0, value=1.0)
with gr.Accordion(label="Screenshot",open=False):
with gr.Row():
with gr.Column(scale=3):
im_btn=gr.Button("Screenshot")
img=gr.Image(type='filepath')
with gr.Column(scale=1):
with gr.Row():
im_height=gr.Number(label="Height",value=5000)
im_width=gr.Number(label="Width",value=500)
wait_time=gr.Number(label="Wait Time",value=3000)
theme=gr.Radio(label="Theme", choices=["light","dark"],value="light")
chatblock=gr.Dropdown(label="Chatblocks",info="Choose specific blocks of chat",choices=[c for c in range(1,40)],multiselect=True)
hid1=gr.Number(value=1)
hid2=gr.Number(value=2)
hid3=gr.Number(value=3)
hid4=gr.Number(value=4)
client_choice.change(load_models,client_choice,[chat_a,chat_b,chat_c,chat_d])
#im_go=im_btn.click(get_screenshot,[chat_b,im_height,im_width,chatblock,theme,wait_time],img)
#chat_sub=inp.submit(check_rand,[rand,seed],seed).then(chat_inf,[sys_inp,inp,chat_b,client_choice,seed,temp,tokens,top_p,rep_p],chat_b)
go1=btn.click(check_rand,[rand,seed],seed).then(chat_inf,[sys_inp,inp,chat_b,client_choice,seed,temp,tokens,top_p,rep_p,hid1],chat_a)
go2=btn.click(check_rand,[rand,seed],seed).then(chat_inf,[sys_inp,inp,chat_b,client_choice,seed,temp,tokens,top_p,rep_p,hid2],chat_b)
go3=btn.click(check_rand,[rand,seed],seed).then(chat_inf,[sys_inp,inp,chat_b,client_choice,seed,temp,tokens,top_p,rep_p,hid3],chat_c)
go4=btn.click(check_rand,[rand,seed],seed).then(chat_inf,[sys_inp,inp,chat_b,client_choice,seed,temp,tokens,top_p,rep_p,hid4],chat_d)
#stop_btn.click(None,None,None,cancels=[go,im_go,chat_sub])
#clear_btn.click(clear_fn,None,[inp,sys_inp,chat_b])
app.queue(default_concurrency_limit=10).launch()
|