Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,12 +1,20 @@
|
|
1 |
import gradio as gr
|
2 |
-
|
3 |
-
from
|
4 |
-
|
5 |
|
6 |
tokenizer = T5TokenizerFast.from_pretrained("t5-base")
|
7 |
-
best_model = PoemSummaryModel.load_from_checkpoint("best-checkpoint.ckpt")
|
8 |
-
best_model.freeze()
|
9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
def encode_text(text):
|
12 |
encoding = tokenizer.encode_plus(
|
@@ -21,7 +29,7 @@ def encode_text(text):
|
|
21 |
|
22 |
def generate_summary(input_ids, attention_mask, model):
|
23 |
model = model.to(input_ids.device)
|
24 |
-
generated_ids = model.
|
25 |
input_ids=input_ids,
|
26 |
attention_mask=attention_mask,
|
27 |
max_length=150,
|
@@ -39,7 +47,7 @@ def decode_summary(generated_ids):
|
|
39 |
|
40 |
def summarize(text):
|
41 |
input_ids, attention_mask = encode_text(text)
|
42 |
-
generated_ids = generate_summary(input_ids, attention_mask,
|
43 |
summary = decode_summary(generated_ids)
|
44 |
return summary
|
45 |
|
@@ -53,4 +61,4 @@ gr.Interface(
|
|
53 |
outputs=output_text,
|
54 |
title="Poem Pulse",
|
55 |
description="Enter a Poem and get its Jist."
|
56 |
-
).launch()
|
|
|
1 |
import gradio as gr
|
2 |
+
import torch
|
3 |
+
from transformers import T5ForConditionalGeneration, T5TokenizerFast
|
|
|
4 |
|
5 |
tokenizer = T5TokenizerFast.from_pretrained("t5-base")
|
|
|
|
|
6 |
|
7 |
+
# Define the quantized model architecture
|
8 |
+
quantized_model = T5ForConditionalGeneration.from_pretrained("t5-base")
|
9 |
+
|
10 |
+
# Load the state dictionary
|
11 |
+
state_dict = torch.load("quantized_model.pt")
|
12 |
+
|
13 |
+
# Filter out keys that are not present in the quantized model
|
14 |
+
filtered_state_dict = {k: v for k, v in state_dict.items() if k in quantized_model.state_dict()}
|
15 |
+
|
16 |
+
# Load the filtered state dictionary into the quantized model
|
17 |
+
quantized_model.load_state_dict(filtered_state_dict, strict=False)
|
18 |
|
19 |
def encode_text(text):
|
20 |
encoding = tokenizer.encode_plus(
|
|
|
29 |
|
30 |
def generate_summary(input_ids, attention_mask, model):
|
31 |
model = model.to(input_ids.device)
|
32 |
+
generated_ids = model.generate(
|
33 |
input_ids=input_ids,
|
34 |
attention_mask=attention_mask,
|
35 |
max_length=150,
|
|
|
47 |
|
48 |
def summarize(text):
|
49 |
input_ids, attention_mask = encode_text(text)
|
50 |
+
generated_ids = generate_summary(input_ids, attention_mask, quantized_model)
|
51 |
summary = decode_summary(generated_ids)
|
52 |
return summary
|
53 |
|
|
|
61 |
outputs=output_text,
|
62 |
title="Poem Pulse",
|
63 |
description="Enter a Poem and get its Jist."
|
64 |
+
).launch()
|