Spaces:
Sleeping
Sleeping
File size: 6,576 Bytes
c6e5fcf c5e92b7 b20ca05 c5e92b7 b20ca05 c5e92b7 c6e5fcf c5e92b7 |
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 |
##########################
###########################
### Inspired from ######
### https://ai.plainenglish.io/run-any-quantized-model-online-for-free-on-hugging-face-a3b9424a441e
############################
import gradio as gr
from llama_cpp import Llama
import datetime
import os
import datetime
from huggingface_hub import hf_hub_download
#MODEL SETTINGS also for DISPLAY
convHistory = ''
modelfile = hf_hub_download(
repo_id=os.environ.get("REPO_ID", "TheBloke/stablelm-zephyr-3b-GGUF"),
filename=os.environ.get("MODEL_FILE", "stablelm-zephyr-3b.Q4_K_M.gguf"),
)
repetitionpenalty = 1.15
contextlength=4096
logfile = 'StableZephyr3b_logs.txt'
print("loading model...")
stt = datetime.datetime.now()
# Set gpu_layers to the number of layers to offload to GPU. Set to 0 if no GPU acceleration is available on your system.
llm = Llama(
model_path=modelfile, # Download the model file first
n_ctx=contextlength, # The max sequence length to use - note that longer sequence lengths require much more resources
#n_threads=2, # The number of CPU threads to use, tailor to your system and the resulting performance
)
dt = datetime.datetime.now() - stt
print(f"Model loaded in {dt}")
def writehistory(text):
with open(logfile, 'a') as f:
f.write(text)
f.write('\n')
f.close()
"""
gr.themes.Base()
gr.themes.Default()
gr.themes.Glass()
gr.themes.Monochrome()
gr.themes.Soft()
"""
def combine(a, b, c, d,e,f):
global convHistory
import datetime
SYSTEM_PROMPT = f"""{a}
"""
temperature = c
max_new_tokens = d
repeat_penalty = f
top_p = e
prompt = f"<|user|>\n{b}<|endoftext|>\n<|assistant|>"
start = datetime.datetime.now()
generation = ""
delta = ""
prompt_tokens = f"Prompt Tokens: {len(llm.tokenize(bytes(prompt,encoding='utf-8')))}"
generated_text = ""
answer_tokens = ''
total_tokens = ''
for character in llm(prompt,
max_tokens=max_new_tokens,
stop=["</s>"],
temperature = temperature,
repeat_penalty = repeat_penalty,
top_p = top_p, # Example stop token - not necessarily correct for this specific model! Please check before using.
echo=False,
stream=True):
generation += character["choices"][0]["text"]
answer_tokens = f"Out Tkns: {len(llm.tokenize(bytes(generation,encoding='utf-8')))}"
total_tokens = f"Total Tkns: {len(llm.tokenize(bytes(prompt,encoding='utf-8'))) + len(llm.tokenize(bytes(generation,encoding='utf-8')))}"
delta = datetime.datetime.now() - start
yield generation, delta, prompt_tokens, answer_tokens, total_tokens
timestamp = datetime.datetime.now()
logger = f"""time: {timestamp}\n Temp: {temperature} - MaxNewTokens: {max_new_tokens} - RepPenalty: 1.5 \nPROMPT: \n{prompt}\nStableZephyr3B: {generation}\nGenerated in {delta}\nPromptTokens: {prompt_tokens} Output Tokens: {answer_tokens} Total Tokens: {total_tokens}\n\n---\n\n"""
writehistory(logger)
convHistory = convHistory + prompt + "\n" + generation + "\n"
print(convHistory)
return generation, delta, prompt_tokens, answer_tokens, total_tokens
#return generation, delta
# MAIN GRADIO INTERFACE
with gr.Blocks(theme='Medguy/base2') as demo: #theme=gr.themes.Glass() #theme='remilia/Ghostly'
#TITLE SECTION
with gr.Row(variant='compact'):
with gr.Column(scale=3):
gr.Image(value='https://github.com/fabiomatricardi/StableLM-Zephyr3B_Playground/raw/main/logo-banner-StableZephyr.jpg',
show_label = False,
show_download_button = False, container = False)
with gr.Column(scale=10):
gr.HTML("<center>"
+ "<h3>Prompt Engineering Playground!</h3>"
+ "<h1>💎🦜 StableLM-Zephyr-3B - 4K context window</h2></center>")
with gr.Row():
with gr.Column(min_width=80):
gentime = gr.Textbox(value="", placeholder="Generation Time:", min_width=50, show_label=False)
with gr.Column(min_width=80):
prompttokens = gr.Textbox(value="", placeholder="Prompt Tkn:", min_width=50, show_label=False)
with gr.Column(min_width=80):
outputokens = gr.Textbox(value="", placeholder="Output Tkn:", min_width=50, show_label=False)
with gr.Column(min_width=80):
totaltokens = gr.Textbox(value="", placeholder="Total Tokens:", min_width=50, show_label=False)
# INTERACTIVE INFOGRAPHIC SECTION
# PLAYGROUND INTERFACE SECTION
with gr.Row():
with gr.Column(scale=1):
gr.Markdown(
f"""
### Tunning Parameters""")
temp = gr.Slider(label="Temperature",minimum=0.0, maximum=1.0, step=0.01, value=0.42)
top_p = gr.Slider(label="Top_P",minimum=0.0, maximum=1.0, step=0.01, value=0.8)
repPen = gr.Slider(label="Repetition Penalty",minimum=0.0, maximum=4.0, step=0.01, value=1.2)
max_len = gr.Slider(label="Maximum output lenght", minimum=10,maximum=(contextlength-500),step=2, value=900)
gr.Markdown(
"""
Fill the System Prompt and User Prompt
And then click the Button below
""")
btn = gr.Button(value="💎🦜 Generate", variant='primary')
gr.Markdown(
f"""
- **Prompt Template**: StableLM-Zephyr 💎🦜
- **Repetition Penalty**: {repetitionpenalty}
- **Context Lenght**: {contextlength} tokens
- **LLM Engine**: llama-cpp
- **Model**: 💎🦜 StableLM-Zephyr-7b
- **Log File**: {logfile}
""")
with gr.Column(scale=4):
txt = gr.Textbox(label="System Prompt", value = "", placeholder = "This models does not have any System prompt...",lines=1, interactive = False)
txt_2 = gr.Textbox(label="User Prompt", lines=6, show_copy_button=True)
txt_3 = gr.Textbox(value="", label="Output", lines = 12, show_copy_button=True)
btn.click(combine, inputs=[txt, txt_2,temp,max_len,top_p,repPen], outputs=[txt_3,gentime,prompttokens,outputokens,totaltokens])
if __name__ == "__main__":
demo.launch(inbrowser=True) |