Update app.py
Browse files
app.py
CHANGED
@@ -30,15 +30,15 @@ model.eval()
|
|
30 |
os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:50'
|
31 |
|
32 |
class StarlingBot:
|
33 |
-
def __init__(self,
|
34 |
self.system_prompt = system_prompt
|
35 |
|
36 |
-
def predict(self, user_message, assistant_message,
|
37 |
try:
|
38 |
if mode == "Assistant":
|
39 |
-
conversation = f"
|
40 |
else: # mode == "Coder"
|
41 |
-
conversation = f"
|
42 |
input_ids = tokenizer.encode(conversation, return_tensors="pt", add_special_tokens=True)
|
43 |
input_ids = input_ids.to(device)
|
44 |
response = model.generate(
|
@@ -72,10 +72,10 @@ examples = [
|
|
72 |
1.9, # repetition_penalty
|
73 |
]
|
74 |
]
|
75 |
-
|
76 |
starling_bot = StarlingBot()
|
77 |
|
78 |
-
def gradio_starling(user_message, assistant_message,
|
79 |
response = starling_bot.predict(user_message, assistant_message, system_prompt, mode, do_sample, temperature, max_new_tokens, top_p, repetition_penalty)
|
80 |
return response
|
81 |
|
@@ -83,8 +83,7 @@ with gr.Blocks(theme="ParityError/Anime") as demo:
|
|
83 |
gr.Markdown(title)
|
84 |
gr.Markdown(description)
|
85 |
with gr.Row():
|
86 |
-
|
87 |
-
assistant_message = gr.Textbox(label="💫🌠Starling Assistant Message", lines=2)
|
88 |
user_message = gr.Textbox(label="Your Message", lines=3)
|
89 |
with gr.Row():
|
90 |
mode = gr.Radio(choices=["Assistant", "Coder"], value="Assistant", label="Mode")
|
@@ -101,7 +100,7 @@ with gr.Blocks(theme="ParityError/Anime") as demo:
|
|
101 |
|
102 |
submit_button.click(
|
103 |
gradio_starling,
|
104 |
-
inputs=[user_message, assistant_message,
|
105 |
outputs=output_text
|
106 |
)
|
107 |
|
|
|
30 |
os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'max_split_size_mb:50'
|
31 |
|
32 |
class StarlingBot:
|
33 |
+
def __init__(self, assistant_message="I am Starling-7B by Tonic-AI, I am ready to do anything to help my user."):
|
34 |
self.system_prompt = system_prompt
|
35 |
|
36 |
+
def predict(self, user_message, assistant_message, mode, do_sample, temperature=0.4, max_new_tokens=700, top_p=0.99, repetition_penalty=1.9):
|
37 |
try:
|
38 |
if mode == "Assistant":
|
39 |
+
conversation = f"GPT4 Correct Assistant: {assistant_message if assistant_message else ''} GPT4 Correct User: {user_message} GPT4 Correct Assistant:"
|
40 |
else: # mode == "Coder"
|
41 |
+
conversation = f"Code Assistant: {assistant_message if assistant_message else ''} Code User:: {user_message} Code Assistant:"
|
42 |
input_ids = tokenizer.encode(conversation, return_tensors="pt", add_special_tokens=True)
|
43 |
input_ids = input_ids.to(device)
|
44 |
response = model.generate(
|
|
|
72 |
1.9, # repetition_penalty
|
73 |
]
|
74 |
]
|
75 |
+
|
76 |
starling_bot = StarlingBot()
|
77 |
|
78 |
+
def gradio_starling(user_message, assistant_message, mode, do_sample, temperature, max_new_tokens, top_p, repetition_penalty):
|
79 |
response = starling_bot.predict(user_message, assistant_message, system_prompt, mode, do_sample, temperature, max_new_tokens, top_p, repetition_penalty)
|
80 |
return response
|
81 |
|
|
|
83 |
gr.Markdown(title)
|
84 |
gr.Markdown(description)
|
85 |
with gr.Row():
|
86 |
+
assistant_message = gr.Textbox(label="Optional💫🌠Starling Assistant Message", lines=2)
|
|
|
87 |
user_message = gr.Textbox(label="Your Message", lines=3)
|
88 |
with gr.Row():
|
89 |
mode = gr.Radio(choices=["Assistant", "Coder"], value="Assistant", label="Mode")
|
|
|
100 |
|
101 |
submit_button.click(
|
102 |
gradio_starling,
|
103 |
+
inputs=[user_message, assistant_message, mode, do_sample, temperature, max_new_tokens, top_p, repetition_penalty],
|
104 |
outputs=output_text
|
105 |
)
|
106 |
|