CreitinGameplays
commited on
Commit
•
7f42dfb
1
Parent(s):
ae647d1
Update app.py
Browse files
app.py
CHANGED
@@ -5,12 +5,12 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
5 |
# Define the BLOOM model name
|
6 |
model_name = "CreitinGameplays/bloom-3b-conversational"
|
7 |
|
8 |
-
# Load tokenizer and model
|
9 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
10 |
model = AutoModelForCausalLM.from_pretrained(model_name)
|
11 |
|
12 |
def generate_text(user_prompt):
|
13 |
-
"""Generates text using the BLOOM model
|
14 |
# Construct the full prompt with system introduction, user prompt, and assistant role
|
15 |
prompt = f"<|system|> You are a helpful AI assistant. </s> <|prompter|> {user_prompt} </s> <|assistant|>"
|
16 |
|
@@ -24,7 +24,7 @@ def generate_text(user_prompt):
|
|
24 |
num_beams=1,
|
25 |
num_return_sequences=1,
|
26 |
do_sample=True,
|
27 |
-
top_k=
|
28 |
top_p=0.95,
|
29 |
temperature=0.2,
|
30 |
repetition_penalty=1.155
|
@@ -40,7 +40,7 @@ def generate_text(user_prompt):
|
|
40 |
|
41 |
return assistant_response
|
42 |
|
43 |
-
# Define the Gradio interface
|
44 |
interface = gr.Interface(
|
45 |
fn=generate_text,
|
46 |
inputs=[
|
@@ -48,6 +48,7 @@ interface = gr.Interface(
|
|
48 |
],
|
49 |
outputs="text",
|
50 |
description="Interact with BLOOM-3b-conversational (Loaded with Hugging Face Transformers)",
|
|
|
51 |
)
|
52 |
|
53 |
# Launch the Gradio interface
|
|
|
5 |
# Define the BLOOM model name
|
6 |
model_name = "CreitinGameplays/bloom-3b-conversational"
|
7 |
|
8 |
+
# Load tokenizer and model (outside the function for efficiency)
|
9 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
10 |
model = AutoModelForCausalLM.from_pretrained(model_name)
|
11 |
|
12 |
def generate_text(user_prompt):
|
13 |
+
"""Generates text using the pre-loaded BLOOM model and removes the user prompt."""
|
14 |
# Construct the full prompt with system introduction, user prompt, and assistant role
|
15 |
prompt = f"<|system|> You are a helpful AI assistant. </s> <|prompter|> {user_prompt} </s> <|assistant|>"
|
16 |
|
|
|
24 |
num_beams=1,
|
25 |
num_return_sequences=1,
|
26 |
do_sample=True,
|
27 |
+
top_k=50,
|
28 |
top_p=0.95,
|
29 |
temperature=0.2,
|
30 |
repetition_penalty=1.155
|
|
|
40 |
|
41 |
return assistant_response
|
42 |
|
43 |
+
# Define the Gradio interface with live=True for real-time updates
|
44 |
interface = gr.Interface(
|
45 |
fn=generate_text,
|
46 |
inputs=[
|
|
|
48 |
],
|
49 |
outputs="text",
|
50 |
description="Interact with BLOOM-3b-conversational (Loaded with Hugging Face Transformers)",
|
51 |
+
live=True
|
52 |
)
|
53 |
|
54 |
# Launch the Gradio interface
|