Spaces:
Runtime error
Runtime error
Stefan Dumitrescu
commited on
Commit
β’
0957f7e
1
Parent(s):
76d0859
Update
Browse files
app.py
CHANGED
@@ -13,22 +13,35 @@ from transformers import AutoTokenizer, AutoModelWithLMHead
|
|
13 |
st.set_page_config(
|
14 |
page_title="Romanian Text Generator",
|
15 |
page_icon="π·π΄",
|
|
|
16 |
)
|
17 |
|
18 |
-
|
|
|
|
|
|
|
19 |
|
20 |
|
21 |
-
@st.cache
|
22 |
-
def load_model(model_name):
|
23 |
-
model = AutoModelWithLMHead.from_pretrained("dumitrescustefan/gpt-neo-romanian-780m")
|
24 |
-
return model
|
25 |
|
26 |
-
model = load_model("dumitrescustefan/gpt-neo-romanian-780m")
|
27 |
|
28 |
|
29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
output_sequences = model.generate(
|
31 |
-
input_ids=input_ids,
|
32 |
max_length=max_length,
|
33 |
temperature=temperature,
|
34 |
top_k=top_k,
|
@@ -40,26 +53,7 @@ def infer(input_ids, max_length, temperature, top_k, top_p):
|
|
40 |
return output_sequences
|
41 |
|
42 |
|
43 |
-
|
44 |
-
|
45 |
-
# prompts
|
46 |
-
st.title("Write")
|
47 |
-
st.write(
|
48 |
-
"The almighty king of text generation, GPT-2 comes in four available sizes, only three of which have been publicly made available. Feared for its fake news generation capabilities, it currently stands as the most syntactically coherent model. A direct successor to the original GPT, it reinforces the already established pre-training/fine-tuning killer duo. From the paper: Language Models are Unsupervised Multitask Learners by Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei and Ilya Sutskever.")
|
49 |
-
|
50 |
-
sent = st.text_area("Text", default_value, height=275)
|
51 |
-
max_length = st.sidebar.slider("Max Length", min_value=10, max_value=30)
|
52 |
-
temperature = st.sidebar.slider("Temperature", value=1.0, min_value=0.0, max_value=1.0, step=0.05)
|
53 |
-
top_k = st.sidebar.slider("Top-k", min_value=0, max_value=5, value=0)
|
54 |
-
top_p = st.sidebar.slider("Top-p", min_value=0.0, max_value=1.0, step=0.05, value=0.9)
|
55 |
-
|
56 |
-
encoded_prompt = tokenizer.encode(sent, add_special_tokens=False, return_tensors="pt")
|
57 |
-
if encoded_prompt.size()[-1] == 0:
|
58 |
-
input_ids = None
|
59 |
-
else:
|
60 |
-
input_ids = encoded_prompt
|
61 |
-
|
62 |
-
output_sequences = infer(input_ids, max_length, temperature, top_k, top_p)
|
63 |
|
64 |
for generated_sequence_idx, generated_sequence in enumerate(output_sequences):
|
65 |
print(f"=== GENERATED SEQUENCE {generated_sequence_idx + 1} ===")
|
@@ -79,5 +73,5 @@ for generated_sequence_idx, generated_sequence in enumerate(output_sequences):
|
|
79 |
generated_sequences.append(total_sequence)
|
80 |
print(total_sequence)
|
81 |
|
82 |
-
st.write(generated_sequences[-1])
|
83 |
|
|
|
13 |
st.set_page_config(
|
14 |
page_title="Romanian Text Generator",
|
15 |
page_icon="π·π΄",
|
16 |
+
layout="wide"
|
17 |
)
|
18 |
|
19 |
+
model_list = ["dumitrescustefan/gpt-neo-romanian-780m"]
|
20 |
+
st.sidebar.header("Select Model")
|
21 |
+
model_checkpoint = st.sidebar.radio("", model_list)
|
22 |
+
text_element = st.text_input('Text:', 'Acesta este un exemplu,')
|
23 |
|
24 |
|
|
|
|
|
|
|
|
|
25 |
|
|
|
26 |
|
27 |
|
28 |
+
st.sidebar.header("Select type of PERSON detection")
|
29 |
+
max_length = st.sidebar.slider("Max Length", value=20, min_value=10, max_value=200)
|
30 |
+
temperature = st.sidebar.slider("Temperature", value=1.0, min_value=0.0, max_value=1.0, step=0.05)
|
31 |
+
top_k = st.sidebar.slider("Top-k", min_value=0, max_value=15, step=1, value=0)
|
32 |
+
top_p = st.sidebar.slider("Top-p", min_value=0.0, max_value=1.0, step=0.05, value=0.9)
|
33 |
+
|
34 |
+
|
35 |
+
@st.cache(allow_output_mutation=True)
|
36 |
+
def setModel(model_name):
|
37 |
+
model = AutoModelWithLMHead.from_pretrained(model_name)
|
38 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
39 |
+
return model, tokenizer
|
40 |
+
|
41 |
+
def infer(model, tokenizer, text, input_ids, max_length, temperature, top_k, top_p):
|
42 |
+
encoded_prompt = tokenizer.encode(text, add_special_tokens=False, return_tensors="pt")
|
43 |
output_sequences = model.generate(
|
44 |
+
input_ids=encoded_prompt.input_ids,
|
45 |
max_length=max_length,
|
46 |
temperature=temperature,
|
47 |
top_k=top_k,
|
|
|
53 |
return output_sequences
|
54 |
|
55 |
|
56 |
+
output_sequences = infer(model, tokenizer, text_element, input_ids, max_length, temperature, top_k, top_p)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
|
58 |
for generated_sequence_idx, generated_sequence in enumerate(output_sequences):
|
59 |
print(f"=== GENERATED SEQUENCE {generated_sequence_idx + 1} ===")
|
|
|
73 |
generated_sequences.append(total_sequence)
|
74 |
print(total_sequence)
|
75 |
|
76 |
+
st.write(generated_sequences[-1], text_element)
|
77 |
|