Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -10,18 +10,17 @@ description = """## Compare Creative Writing: Standard Sampler vs. Backtrack Sam
|
|
10 |
This is a demo of the [Backtrack Sampler](https://github.com/Mihaiii/backtrack_sampler) framework using "Creative Writing Strategy".
|
11 |
<br />On the left you have the output of the standard sampling and on the write the output privided by Backtrack Sampler.
|
12 |
"""
|
13 |
-
|
14 |
model_name = "unsloth/Llama-3.2-1B-Instruct"
|
|
|
15 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
16 |
|
17 |
-
# Load two instances of the model on CUDA for parallel inference
|
18 |
model = AutoModelForCausalLM.from_pretrained(model_name).to("cuda")
|
19 |
|
20 |
provider = TransformersProvider(model, tokenizer, device)
|
21 |
strategy = CreativeWritingStrategy(provider)
|
22 |
creative_sampler = BacktrackSampler(strategy, provider)
|
23 |
|
24 |
-
# Helper function to create message array for the chat template
|
25 |
def create_chat_template_messages(history, prompt):
|
26 |
messages = [{"role": "user", "content": prompt}]
|
27 |
|
@@ -33,14 +32,11 @@ def create_chat_template_messages(history, prompt):
|
|
33 |
|
34 |
@spaces.GPU
|
35 |
def generate_responses(prompt, history):
|
36 |
-
# Create messages array for chat history and apply template
|
37 |
messages = create_chat_template_messages(history, prompt)
|
38 |
wrapped_prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_special_tokens=True, add_generation_prompt=True)
|
39 |
|
40 |
-
#already has special tokens
|
41 |
inputs = tokenizer.encode(wrapped_prompt, add_special_tokens=False, return_tensors="pt").to("cuda")
|
42 |
|
43 |
-
# Custom sampler task: loop over generator and collect outputs in a list
|
44 |
async def custom_sampler_task():
|
45 |
generated_list = []
|
46 |
generator = creative_sampler.generate(wrapped_prompt, max_length=2048, temperature=1)
|
@@ -50,51 +46,39 @@ def generate_responses(prompt, history):
|
|
50 |
|
51 |
custom_output = asyncio.run(custom_sampler_task())
|
52 |
standard_output = model.generate(inputs, max_length=2048, temperature=1)
|
53 |
-
# Decode standard output and remove the prompt from the generated response
|
54 |
standard_response = tokenizer.decode(standard_output[0][len(inputs[0]):], skip_special_tokens=True)
|
55 |
|
56 |
return standard_response.strip(), custom_output.strip()
|
57 |
|
58 |
-
# Create the Gradio interface with the Citrus theme
|
59 |
with gr.Blocks(theme=gr.themes.Citrus()) as demo:
|
60 |
gr.Markdown(description)
|
61 |
|
62 |
-
# Chatbot components
|
63 |
with gr.Row():
|
64 |
standard_chat = gr.Chatbot(label="Standard Sampler")
|
65 |
custom_chat = gr.Chatbot(label="Creative Writing Strategy")
|
66 |
|
67 |
-
# Input components
|
68 |
with gr.Row():
|
69 |
prompt_input = gr.Textbox(label="Enter your prompt", placeholder="Type your message here...", lines=1)
|
70 |
|
71 |
-
# Example prompts
|
72 |
examples = [
|
73 |
"Write me a short story about a talking dog who wants to be a detective.",
|
74 |
"Tell me a short tale of a dragon who is afraid of heights.",
|
75 |
"Create a short story where aliens land on Earth, but they just want to throw a party."
|
76 |
]
|
77 |
|
78 |
-
# Add example buttons
|
79 |
gr.Examples(examples=examples, inputs=prompt_input)
|
80 |
|
81 |
-
# Button to submit the prompt
|
82 |
submit_button = gr.Button("Submit")
|
83 |
|
84 |
-
# Function to handle chat updates
|
85 |
def update_chat(prompt, standard_history, custom_history):
|
86 |
standard_response, custom_response = generate_responses(prompt, standard_history)
|
87 |
|
88 |
-
# Append new responses to chat histories
|
89 |
standard_history = standard_history + [(prompt, standard_response)]
|
90 |
custom_history = custom_history + [(prompt, custom_response)]
|
91 |
|
92 |
-
# Clear the input field after submission
|
93 |
return standard_history, custom_history, ""
|
94 |
|
95 |
-
# Bind the submit button to the update function and allow pressing Enter to submit
|
96 |
prompt_input.submit(fn=update_chat, inputs=[prompt_input, standard_chat, custom_chat], outputs=[standard_chat, custom_chat, prompt_input])
|
97 |
submit_button.click(fn=update_chat, inputs=[prompt_input, standard_chat, custom_chat], outputs=[standard_chat, custom_chat, prompt_input])
|
98 |
|
99 |
-
# Launch the app with queueing and sharing enabled
|
100 |
demo.queue().launch(debug=True)
|
|
|
10 |
This is a demo of the [Backtrack Sampler](https://github.com/Mihaiii/backtrack_sampler) framework using "Creative Writing Strategy".
|
11 |
<br />On the left you have the output of the standard sampling and on the write the output privided by Backtrack Sampler.
|
12 |
"""
|
13 |
+
|
14 |
model_name = "unsloth/Llama-3.2-1B-Instruct"
|
15 |
+
device = torch.device('cuda')
|
16 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
17 |
|
|
|
18 |
model = AutoModelForCausalLM.from_pretrained(model_name).to("cuda")
|
19 |
|
20 |
provider = TransformersProvider(model, tokenizer, device)
|
21 |
strategy = CreativeWritingStrategy(provider)
|
22 |
creative_sampler = BacktrackSampler(strategy, provider)
|
23 |
|
|
|
24 |
def create_chat_template_messages(history, prompt):
|
25 |
messages = [{"role": "user", "content": prompt}]
|
26 |
|
|
|
32 |
|
33 |
@spaces.GPU
|
34 |
def generate_responses(prompt, history):
|
|
|
35 |
messages = create_chat_template_messages(history, prompt)
|
36 |
wrapped_prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_special_tokens=True, add_generation_prompt=True)
|
37 |
|
|
|
38 |
inputs = tokenizer.encode(wrapped_prompt, add_special_tokens=False, return_tensors="pt").to("cuda")
|
39 |
|
|
|
40 |
async def custom_sampler_task():
|
41 |
generated_list = []
|
42 |
generator = creative_sampler.generate(wrapped_prompt, max_length=2048, temperature=1)
|
|
|
46 |
|
47 |
custom_output = asyncio.run(custom_sampler_task())
|
48 |
standard_output = model.generate(inputs, max_length=2048, temperature=1)
|
|
|
49 |
standard_response = tokenizer.decode(standard_output[0][len(inputs[0]):], skip_special_tokens=True)
|
50 |
|
51 |
return standard_response.strip(), custom_output.strip()
|
52 |
|
|
|
53 |
with gr.Blocks(theme=gr.themes.Citrus()) as demo:
|
54 |
gr.Markdown(description)
|
55 |
|
|
|
56 |
with gr.Row():
|
57 |
standard_chat = gr.Chatbot(label="Standard Sampler")
|
58 |
custom_chat = gr.Chatbot(label="Creative Writing Strategy")
|
59 |
|
|
|
60 |
with gr.Row():
|
61 |
prompt_input = gr.Textbox(label="Enter your prompt", placeholder="Type your message here...", lines=1)
|
62 |
|
|
|
63 |
examples = [
|
64 |
"Write me a short story about a talking dog who wants to be a detective.",
|
65 |
"Tell me a short tale of a dragon who is afraid of heights.",
|
66 |
"Create a short story where aliens land on Earth, but they just want to throw a party."
|
67 |
]
|
68 |
|
|
|
69 |
gr.Examples(examples=examples, inputs=prompt_input)
|
70 |
|
|
|
71 |
submit_button = gr.Button("Submit")
|
72 |
|
|
|
73 |
def update_chat(prompt, standard_history, custom_history):
|
74 |
standard_response, custom_response = generate_responses(prompt, standard_history)
|
75 |
|
|
|
76 |
standard_history = standard_history + [(prompt, standard_response)]
|
77 |
custom_history = custom_history + [(prompt, custom_response)]
|
78 |
|
|
|
79 |
return standard_history, custom_history, ""
|
80 |
|
|
|
81 |
prompt_input.submit(fn=update_chat, inputs=[prompt_input, standard_chat, custom_chat], outputs=[standard_chat, custom_chat, prompt_input])
|
82 |
submit_button.click(fn=update_chat, inputs=[prompt_input, standard_chat, custom_chat], outputs=[standard_chat, custom_chat, prompt_input])
|
83 |
|
|
|
84 |
demo.queue().launch(debug=True)
|