Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -2,10 +2,10 @@ import gradio as gr
|
|
2 |
import torch
|
3 |
from mingru_lm import MinGRU_LM
|
4 |
|
5 |
-
|
6 |
model = MinGRU_LM(dim=512, num_tokens=256, num_layers=6)
|
7 |
pt_model = "best_model.pt"
|
8 |
-
checkpoint = torch.load(pt_model,map_location=torch.device('cpu'))
|
9 |
model.load_state_dict(checkpoint['model_state_dict'])
|
10 |
|
11 |
# Move model to GPU if available
|
@@ -25,36 +25,52 @@ def generate_text(start_text, max_length, temperature):
|
|
25 |
input_tensor = torch.tensor(tokens, dtype=torch.long).unsqueeze(0).to(device) # Ensure long tensor
|
26 |
|
27 |
generated_tokens = tokens.copy()
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
|
|
32 |
|
33 |
last_token_logits = logits[0, -1, :] / temperature
|
34 |
probs = torch.softmax(last_token_logits, dim=-1)
|
|
|
|
|
35 |
next_token = torch.multinomial(probs, num_samples=1).item()
|
36 |
|
37 |
-
# Only append
|
38 |
if next_token < 256:
|
39 |
generated_tokens.append(next_token)
|
40 |
input_tensor = torch.cat([input_tensor, torch.tensor([[next_token]], device=device)], dim=1)
|
|
|
|
|
41 |
else:
|
42 |
continue # Skip tokens outside ASCII range
|
43 |
|
44 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
|
46 |
# Gradio interface
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
gr.
|
52 |
-
gr.Slider(minimum=
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
)
|
58 |
-
|
59 |
-
|
60 |
-
|
|
|
|
|
|
|
|
|
|
|
|
2 |
import torch
|
3 |
from mingru_lm import MinGRU_LM
|
4 |
|
5 |
+
# Load the model
|
6 |
model = MinGRU_LM(dim=512, num_tokens=256, num_layers=6)
|
7 |
pt_model = "best_model.pt"
|
8 |
+
checkpoint = torch.load(pt_model, map_location=torch.device('cpu'))
|
9 |
model.load_state_dict(checkpoint['model_state_dict'])
|
10 |
|
11 |
# Move model to GPU if available
|
|
|
25 |
input_tensor = torch.tensor(tokens, dtype=torch.long).unsqueeze(0).to(device) # Ensure long tensor
|
26 |
|
27 |
generated_tokens = tokens.copy()
|
28 |
+
|
29 |
+
# Use a generator to yield tokens one by one
|
30 |
+
for _ in range(max_length):
|
31 |
+
with torch.no_grad():
|
32 |
+
logits = model(input_tensor, labels=None)[1] # Get logits directly
|
33 |
|
34 |
last_token_logits = logits[0, -1, :] / temperature
|
35 |
probs = torch.softmax(last_token_logits, dim=-1)
|
36 |
+
|
37 |
+
# Sample the next token
|
38 |
next_token = torch.multinomial(probs, num_samples=1).item()
|
39 |
|
40 |
+
# Only append valid tokens
|
41 |
if next_token < 256:
|
42 |
generated_tokens.append(next_token)
|
43 |
input_tensor = torch.cat([input_tensor, torch.tensor([[next_token]], device=device)], dim=1)
|
44 |
+
|
45 |
+
yield decode_tokens(generated_tokens)
|
46 |
else:
|
47 |
continue # Skip tokens outside ASCII range
|
48 |
|
49 |
+
yield decode_tokens(generated_tokens)
|
50 |
+
|
51 |
+
def wrapper_generate_text(start_text, max_length, temperature):
|
52 |
+
async_gen = generate_text(start_text, max_length, temperature)
|
53 |
+
|
54 |
+
for output in async_gen:
|
55 |
+
yield output
|
56 |
|
57 |
# Gradio interface
|
58 |
+
with gr.Blocks() as iface:
|
59 |
+
gr.Markdown("### Please be patient, generating text will take some time...")
|
60 |
+
|
61 |
+
with gr.Row():
|
62 |
+
textbox = gr.Textbox(lines=3, label="Enter your prompt", value="Once upon a time")
|
63 |
+
max_length = gr.Slider(minimum=10, maximum=500, value=200, step=1, label="Max Length")
|
64 |
+
temperature = gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature")
|
65 |
+
|
66 |
+
output_textbox = gr.Textbox(lines=10, label="Generated Text")
|
67 |
+
|
68 |
+
btn = gr.Button("Generate Text")
|
69 |
+
|
70 |
+
btn.click(
|
71 |
+
wrapper_generate_text,
|
72 |
+
inputs=[textbox, max_length, temperature],
|
73 |
+
outputs=output_textbox
|
74 |
+
)
|
75 |
+
|
76 |
+
iface.launch(show_api=False, server_name="0.0.0.0")
|