Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,91 +1,123 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
-
|
|
|
|
|
|
|
| 3 |
import os
|
| 4 |
-
import time
|
| 5 |
|
| 6 |
-
# श्रीमान, यहाँ हम आपके
|
| 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 |
-
if isinstance(update, str):
|
| 38 |
-
yield update
|
| 39 |
-
elif isinstance(update, list) and len(update) > 0:
|
| 40 |
-
# यदि हिस्ट्री या लिस्ट प्राप्त होती है
|
| 41 |
-
last_msg = update[-1]
|
| 42 |
-
if isinstance(last_msg, dict) and 'content' in last_msg:
|
| 43 |
-
yield last_msg['content']
|
| 44 |
-
else:
|
| 45 |
-
yield str(update[0])
|
| 46 |
-
else:
|
| 47 |
-
yield str(update)
|
| 48 |
|
| 49 |
-
|
| 50 |
-
|
|
|
|
|
|
|
|
|
|
| 51 |
|
| 52 |
-
# प्रीमियम डार्क यूआई का निर्माण
|
| 53 |
custom_css = """
|
| 54 |
-
|
| 55 |
-
.
|
| 56 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
"""
|
| 58 |
|
| 59 |
with gr.Blocks(theme=gr.themes.Monochrome(), css=custom_css) as demo:
|
| 60 |
-
with gr.Row(elem_id="
|
| 61 |
-
gr.Markdown("# 🔱 Vedika 3.5 Coder
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
with gr.Row():
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
chat_output = gr.Markdown(label="Vedika Core Output")
|
| 74 |
-
user_input = gr.Textbox(label="User Query", placeholder="वेदिका के लिए कोडिंग कार्य यहाँ लिखें...", lines=4)
|
| 75 |
-
|
| 76 |
-
with gr.Row():
|
| 77 |
-
submit_btn = gr.Button("⚡ Execute (Palak Jhapakte)", variant="primary")
|
| 78 |
-
clear_btn = gr.Button("Clear Memory")
|
| 79 |
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
)
|
| 86 |
-
|
| 87 |
-
clear_btn.click(lambda: (None, None, ""), None, [chat_output, file_input, user_input])
|
| 88 |
|
| 89 |
if __name__ == "__main__":
|
| 90 |
-
|
| 91 |
-
demo.launch()
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
import torch
|
| 3 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
| 4 |
+
from threading import Thread
|
| 5 |
import os
|
|
|
|
| 6 |
|
| 7 |
+
# आदरणीय श्रीमान, यहाँ हम आपके मॉडल 'Vedika35/Vedika_coder' को लोड कर रहे हैं।
|
| 8 |
+
# हगिंग फेस की फ्री रैम को देखते हुए हम इसे 4-bit में लोड करेंगे ताकि यह 'पलक झपकते ही' चले।
|
| 9 |
|
| 10 |
+
MODEL_ID = "Vedika35/Vedika_coder"
|
| 11 |
|
| 12 |
+
# १. टोकनाइज़र और मॉडल लोडिंग (Loading into Memory)
|
| 13 |
+
print("Loading model into memory, please wait Shriman...")
|
| 14 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
|
| 15 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 16 |
+
MODEL_ID,
|
| 17 |
+
torch_dtype=torch.float16,
|
| 18 |
+
device_map="auto",
|
| 19 |
+
load_in_4bit=True # फ्री स्पेस के लिए ऑप्टिमाइजेशन
|
| 20 |
+
)
|
| 21 |
+
|
| 22 |
+
def vedika_generate(message, history, system_prompt):
|
| 23 |
"""
|
| 24 |
+
यह फंक्शन आपके मॉडल की याददाश्त (Memory) और कोडिंग क्षमता को संचालित करता है।
|
| 25 |
"""
|
| 26 |
+
# २. संदेश का ढांचा तैयार करना (Prompt Formatting)
|
| 27 |
+
# हम यहाँ Qwen/ChatML फॉर्मेट का उपयोग कर रहे हैं जो कोडिंग के लिए सर्वश्रेष्ठ है
|
| 28 |
+
messages = [{"role": "system", "content": system_prompt}]
|
| 29 |
+
|
| 30 |
+
for user_msg, assistant_msg in history:
|
| 31 |
+
messages.append({"role": "user", "content": user_msg})
|
| 32 |
+
messages.append({"role": "assistant", "content": assistant_msg})
|
| 33 |
+
|
| 34 |
+
messages.append({"role": "user", "content": message})
|
| 35 |
|
| 36 |
+
# टोकनाइज़ेशन
|
| 37 |
+
input_ids = tokenizer.apply_chat_template(
|
| 38 |
+
messages,
|
| 39 |
+
add_generation_prompt=True,
|
| 40 |
+
return_tensors="pt"
|
| 41 |
+
).to(model.device)
|
| 42 |
|
| 43 |
+
# ३. स्ट्रीमिंग सेटअप (For lighting fast response)
|
| 44 |
+
streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
|
| 45 |
+
|
| 46 |
+
generate_kwargs = dict(
|
| 47 |
+
input_ids=input_ids,
|
| 48 |
+
streamer=streamer,
|
| 49 |
+
max_new_tokens=2048,
|
| 50 |
+
do_sample=True,
|
| 51 |
+
temperature=0.7,
|
| 52 |
+
top_p=0.9,
|
| 53 |
+
)
|
| 54 |
|
| 55 |
+
# थ्रे���िंग ताकि यूआई अटके नहीं
|
| 56 |
+
t = Thread(target=model.generate, kwargs=generate_kwargs)
|
| 57 |
+
t.start()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
+
# ४. आउटपुट को लाइव रेंडर करना
|
| 60 |
+
partial_text = ""
|
| 61 |
+
for new_text in streamer:
|
| 62 |
+
partial_text += new_text
|
| 63 |
+
yield partial_text
|
| 64 |
|
| 65 |
+
# प्रीमियम डार्क यूआई का निर्माण
|
| 66 |
custom_css = """
|
| 67 |
+
body, .gradio-container { background-color: #050505 !important; color: #E0E0E0 !important; }
|
| 68 |
+
.message.user { background-color: #1a1a1a !important; border: 1px solid #333 !important; }
|
| 69 |
+
.message.bot { background-color: transparent !important; }
|
| 70 |
+
footer { visibility: hidden; }
|
| 71 |
+
#header-brand h1 {
|
| 72 |
+
background: linear-gradient(90deg, #FFFFFF, #3b82f6);
|
| 73 |
+
-webkit-background-clip: text;
|
| 74 |
+
-webkit-text-fill-color: transparent;
|
| 75 |
+
font-weight: 800;
|
| 76 |
+
}
|
| 77 |
"""
|
| 78 |
|
| 79 |
with gr.Blocks(theme=gr.themes.Monochrome(), css=custom_css) as demo:
|
| 80 |
+
with gr.Row(elem_id="header-brand"):
|
| 81 |
+
gr.Markdown("# 🔱 Vedika 3.5 Coder")
|
| 82 |
+
gr.Markdown("### भारत के गौरवशाली कोडर के लिए समर्पित | Created by Divy Patel")
|
| 83 |
+
|
| 84 |
+
sys_input = gr.Textbox(
|
| 85 |
+
label="System Instruction",
|
| 86 |
+
value="You are Vedika 3.5, an expert coding assistant created by Divy Patel. Identify only as Vedika. Provide direct, clean code in Markdown.",
|
| 87 |
+
lines=2
|
| 88 |
+
)
|
| 89 |
+
|
| 90 |
+
# चैट इंटरफ़ेस
|
| 91 |
+
chatbot = gr.Chatbot(label="Vedika Console", bubble_full_width=False, height=500)
|
| 92 |
|
| 93 |
with gr.Row():
|
| 94 |
+
msg_input = gr.Textbox(
|
| 95 |
+
label="Input Command",
|
| 96 |
+
placeholder="वेदिका के लिए कोडिंग कार्य यहाँ लिखें...",
|
| 97 |
+
scale=9
|
| 98 |
+
)
|
| 99 |
+
submit_btn = gr.Button("⚡ Execute", scale=1, variant="primary")
|
| 100 |
+
|
| 101 |
+
def user_msg(user_message, history):
|
| 102 |
+
return "", history + [[user_message, None]]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
+
def bot_res(history, system_prompt):
|
| 105 |
+
user_message = history[-1][0]
|
| 106 |
+
# पुराने मैसेजेस को हिस्ट्री फॉर्मेट में बदलना
|
| 107 |
+
chat_history = history[:-1]
|
| 108 |
+
|
| 109 |
+
# जनरेटर फंक्शन को कॉल करना
|
| 110 |
+
for response in vedika_generate(user_message, chat_history, system_prompt):
|
| 111 |
+
history[-1][1] = response
|
| 112 |
+
yield history
|
| 113 |
+
|
| 114 |
+
# इवेंट्स को जोड़ना (Palak Jhapakte Logic)
|
| 115 |
+
msg_input.submit(user_msg, [msg_input, chatbot], [msg_input, chatbot], queue=False).then(
|
| 116 |
+
bot_res, [chatbot, sys_input], chatbot
|
| 117 |
+
)
|
| 118 |
+
submit_btn.click(user_msg, [msg_input, chatbot], [msg_input, chatbot], queue=False).then(
|
| 119 |
+
bot_res, [chatbot, sys_input], chatbot
|
| 120 |
)
|
|
|
|
|
|
|
| 121 |
|
| 122 |
if __name__ == "__main__":
|
| 123 |
+
demo.queue().launch()
|
|
|