Spaces:
Paused
Paused
Update App1.py
Browse files
App1.py
CHANGED
@@ -1,34 +1,80 @@
|
|
1 |
-
import gradio as gr
|
2 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer
|
3 |
import torch
|
|
|
|
|
|
|
|
|
4 |
|
|
|
5 |
MODEL_ID = "goonsai-com/civitaiprompts"
|
6 |
-
MODEL_VARIANT = "Q4_K_M" #
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
|
8 |
-
|
9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
model = AutoModelForCausalLM.from_pretrained(
|
11 |
-
|
12 |
-
|
13 |
-
|
|
|
|
|
14 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
-
|
17 |
-
|
18 |
-
|
|
|
|
|
|
|
|
|
|
|
19 |
**inputs,
|
20 |
-
|
|
|
21 |
temperature=0.7,
|
22 |
-
|
23 |
)
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
-
|
|
|
|
|
|
|
|
1 |
import torch
|
2 |
+
import logging
|
3 |
+
import time
|
4 |
+
import gradio as gr
|
5 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
6 |
|
7 |
+
# ---------------- CONFIG ----------------
|
8 |
MODEL_ID = "goonsai-com/civitaiprompts"
|
9 |
+
MODEL_VARIANT = "Q4_K_M" # This is the HF tag for the quantized model
|
10 |
+
MODEL_NAME = "CivitAI-Prompts-Q4_K_M"
|
11 |
+
|
12 |
+
# ---------------- LOGGING ----------------
|
13 |
+
logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
|
14 |
+
logger = logging.getLogger(__name__)
|
15 |
+
logger.info("Starting Gradio chatbot...")
|
16 |
|
17 |
+
# ---------------- LOAD MODEL ----------------
|
18 |
+
logger.info(f"Loading tokenizer from {MODEL_ID} (revision={MODEL_VARIANT})")
|
19 |
+
tokenizer = AutoTokenizer.from_pretrained(
|
20 |
+
MODEL_ID,
|
21 |
+
revision=MODEL_VARIANT,
|
22 |
+
trust_remote_code=True
|
23 |
+
)
|
24 |
+
|
25 |
+
dtype = torch.float16 if torch.cuda.is_available() else torch.float32
|
26 |
+
logger.info(f"Loading model with dtype {dtype}")
|
27 |
model = AutoModelForCausalLM.from_pretrained(
|
28 |
+
MODEL_ID,
|
29 |
+
revision=MODEL_VARIANT,
|
30 |
+
torch_dtype=dtype,
|
31 |
+
device_map="auto",
|
32 |
+
trust_remote_code=True
|
33 |
)
|
34 |
+
logger.info("Model loaded successfully.")
|
35 |
+
|
36 |
+
# ---------------- CHAT FUNCTION ----------------
|
37 |
+
def chat_fn(message):
|
38 |
+
logger.info(f"Received message: {message}")
|
39 |
+
|
40 |
+
# Build prompt
|
41 |
+
full_text = f"User: {message}\nAssistant:"
|
42 |
+
logger.info(f"Full prompt for generation:\n{full_text}")
|
43 |
|
44 |
+
start_time = time.time()
|
45 |
+
# Tokenize input
|
46 |
+
inputs = tokenizer([full_text], return_tensors="pt", truncation=True, max_length=1024).to(model.device)
|
47 |
+
logger.info("Tokenized input.")
|
48 |
+
|
49 |
+
# Generate response
|
50 |
+
logger.info("Generating response...")
|
51 |
+
reply_ids = model.generate(
|
52 |
**inputs,
|
53 |
+
max_new_tokens=512,
|
54 |
+
do_sample=True,
|
55 |
temperature=0.7,
|
56 |
+
top_p=0.9
|
57 |
)
|
58 |
+
response = tokenizer.batch_decode(reply_ids, skip_special_tokens=True)[0]
|
59 |
+
assistant_reply = response.split("Assistant:")[-1].strip()
|
60 |
+
logger.info(f"Assistant reply: {assistant_reply}")
|
61 |
+
logger.info(f"Generation time: {time.time() - start_time:.2f}s")
|
62 |
+
|
63 |
+
return assistant_reply
|
64 |
+
|
65 |
+
# ---------------- GRADIO BLOCKS UI ----------------
|
66 |
+
with gr.Blocks() as demo:
|
67 |
+
gr.Markdown(f"# 🤖 {MODEL_NAME} (Stateless)")
|
68 |
+
|
69 |
+
with gr.Row():
|
70 |
+
with gr.Column():
|
71 |
+
message = gr.Textbox(label="Type your message...", placeholder="Hello!")
|
72 |
+
send_btn = gr.Button("Send")
|
73 |
+
with gr.Column():
|
74 |
+
output = gr.Textbox(label="Assistant Response", lines=10)
|
75 |
+
|
76 |
+
send_btn.click(chat_fn, inputs=[message], outputs=[output])
|
77 |
+
message.submit(chat_fn, inputs=[message], outputs=[output])
|
78 |
|
79 |
+
logger.info("Launching Gradio app...")
|
80 |
+
demo.launch()
|