Spaces:
Runtime error
Runtime error
juanesvelez
commited on
Commit
•
d742495
1
Parent(s):
98c78fa
Update app.py
Browse files
app.py
CHANGED
@@ -1,42 +1,54 @@
|
|
1 |
import solara
|
|
|
2 |
import torch
|
3 |
import torch.nn.functional as F
|
4 |
import pandas as pd
|
5 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
6 |
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
11 |
-
|
12 |
-
text = solara.reactive("Escribe algo en español")
|
13 |
-
|
14 |
@solara.component
|
15 |
def Page():
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
Page()
|
|
|
1 |
import solara
|
2 |
+
import random
|
3 |
import torch
|
4 |
import torch.nn.functional as F
|
5 |
import pandas as pd
|
6 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
7 |
|
8 |
+
tokenizer = AutoTokenizer.from_pretrained('datificate/gpt2-small-spanish')
|
9 |
+
model = AutoModelForCausalLM.from_pretrained('datificate/gpt2-small-spanish')
|
10 |
+
text1 = solara.reactive("Never gonna give you up, never gonna let you")
|
|
|
|
|
|
|
|
|
11 |
@solara.component
|
12 |
def Page():
|
13 |
+
with solara.Column(margin=10):
|
14 |
+
solara.Markdown("#Next token prediction visualization")
|
15 |
+
solara.Markdown("I built this tool to help me understand autoregressive language models. For any given text, it gives the top 10 candidates to be the next token with their respective probabilities. The language model I'm using is the smallest version of GPT-2, with 124M parameters.")
|
16 |
+
def on_action_cell(column, row_index):
|
17 |
+
text1.value += tokenizer.decode(top_10.indices[0][row_index])
|
18 |
+
cell_actions = [solara.CellAction(icon="mdi-thumb-up", name="Select", on_click=on_action_cell)]
|
19 |
+
solara.InputText("Enter text:", value=text1, continuous_update=True)
|
20 |
+
if text1.value != "":
|
21 |
+
tokens = tokenizer.encode(text1.value, return_tensors="pt")
|
22 |
+
spans1 = ""
|
23 |
+
spans2 = ""
|
24 |
+
for i, token in enumerate(tokens[0]):
|
25 |
+
random.seed(i)
|
26 |
+
random_color = ''.join([random.choice('0123456789ABCDEF') for k in range(6)])
|
27 |
+
spans1 += " " + f"<span style='font-family: helvetica; color: #{random_color}'>{token}</span>"
|
28 |
+
spans2 += " " + f"""<span style="
|
29 |
+
padding: 6px;
|
30 |
+
border-right: 3px solid white;
|
31 |
+
line-height: 3em;
|
32 |
+
font-family: courier;
|
33 |
+
background-color: #{random_color};
|
34 |
+
color: white;
|
35 |
+
position: relative;
|
36 |
+
"><span style="
|
37 |
+
position: absolute;
|
38 |
+
top: 5.5ch;
|
39 |
+
line-height: 1em;
|
40 |
+
left: -0.5px;
|
41 |
+
font-size: 0.45em"> {token}</span>{tokenizer.decode([token])}</span>"""
|
42 |
+
solara.Markdown(f'{spans2}')
|
43 |
+
solara.Markdown(f'{spans1}')
|
44 |
+
outputs = model.generate(tokens, max_new_tokens=1, output_scores=True, return_dict_in_generate=True, pad_token_id=tokenizer.eos_token_id)
|
45 |
+
scores = F.softmax(outputs.scores[0], dim=-1)
|
46 |
+
top_10 = torch.topk(scores, 10)
|
47 |
+
df = pd.DataFrame()
|
48 |
+
df["probs"] = top_10.values[0]
|
49 |
+
df["probs"] = [f"{value:.2%}" for value in df["probs"].values]
|
50 |
+
df["next token ID"] = [top_10.indices[0][i].numpy() for i in range(10)]
|
51 |
+
df["predicted next token"] = [tokenizer.decode(top_10.indices[0][i]) for i in range(10)]
|
52 |
+
solara.Markdown("###Prediction")
|
53 |
+
solara.DataFrame(df, items_per_page=10, cell_actions=cell_actions)
|
54 |
Page()
|