Spaces:
Running
Running
abdiharyadi
commited on
Commit
β’
1556875
1
Parent(s):
35d002b
feat: integrate AMRToText and reactivate TextToAMR
Browse files
app.py
CHANGED
@@ -16,6 +16,7 @@ sys.path.append("./amr-tst-indo")
|
|
16 |
from text_to_amr import TextToAMR
|
17 |
from style_detector import StyleDetector
|
18 |
from style_rewriting import StyleRewriting
|
|
|
19 |
|
20 |
amr_parsing_model_name = "mbart-en-id-smaller-indo-amr-parsing-translated-nafkhan"
|
21 |
snapshot_download(
|
@@ -56,6 +57,20 @@ sr = StyleRewriting(
|
|
56 |
maximize_style_words_expansion=False
|
57 |
)
|
58 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
def run(text, source_style):
|
60 |
yield (
|
61 |
"(Memproses ...)",
|
@@ -67,15 +82,7 @@ def run(text, source_style):
|
|
67 |
|
68 |
start_time = time.time()
|
69 |
|
70 |
-
|
71 |
-
source_amr = penman.decode("""
|
72 |
-
(z0 / dan
|
73 |
-
:op1 (z1 / bagus-01
|
74 |
-
:ARG1 (z2 / tempat)
|
75 |
-
:degree (z3 / sangat))
|
76 |
-
:op2 (z4 / bersih-01
|
77 |
-
:ARG1 z2))
|
78 |
-
""")
|
79 |
source_amr.metadata = {}
|
80 |
source_amr_display = penman.encode(source_amr)
|
81 |
source_amr_display += f"\n\n({time.time() - start_time:.2f} s)"
|
@@ -88,10 +95,6 @@ def run(text, source_style):
|
|
88 |
)
|
89 |
|
90 |
triplets = sd.get_triplets(text)
|
91 |
-
# triplets = [
|
92 |
-
# ("kamar", "sangat bagus", "positif"),
|
93 |
-
# ("kamar", "bersih", "positif")
|
94 |
-
# ]
|
95 |
triplets_display = "\n".join(f"({x[0]}, {x[1]}, {x[2]})" for x in triplets)
|
96 |
triplets_display += f"\n\n({time.time() - start_time:.2f} s)"
|
97 |
yield (
|
@@ -103,7 +106,6 @@ def run(text, source_style):
|
|
103 |
)
|
104 |
|
105 |
style_words = sd.get_style_words_from_triplets(triplets)
|
106 |
-
# style_words = ["bagus", "bersih"]
|
107 |
style_words_display = ", ".join(style_words)
|
108 |
style_words_display += f"\n\n({time.time() - start_time:.2f} s)"
|
109 |
yield (
|
@@ -125,7 +127,7 @@ def run(text, source_style):
|
|
125 |
"(Memproses ...)",
|
126 |
)
|
127 |
|
128 |
-
result =
|
129 |
result += f"\n\n({time.time() - start_time:.2f} s)"
|
130 |
yield (
|
131 |
source_amr_display,
|
|
|
16 |
from text_to_amr import TextToAMR
|
17 |
from style_detector import StyleDetector
|
18 |
from style_rewriting import StyleRewriting
|
19 |
+
from amr_to_text import AMRToTextWithTaufiqMethod
|
20 |
|
21 |
amr_parsing_model_name = "mbart-en-id-smaller-indo-amr-parsing-translated-nafkhan"
|
22 |
snapshot_download(
|
|
|
57 |
maximize_style_words_expansion=False
|
58 |
)
|
59 |
|
60 |
+
amr_gen_model_name = "taufiq-indo-amr-generation-gold-uncased"
|
61 |
+
model_path = f"./{amr_gen_model_name}"
|
62 |
+
snapshot_download(
|
63 |
+
repo_id=f"abdiharyadi/{amr_gen_model_name}",
|
64 |
+
local_dir=model_path,
|
65 |
+
allow_patterns=[
|
66 |
+
"*checkpoint-3*"
|
67 |
+
]
|
68 |
+
)
|
69 |
+
a2t = AMRToTextWithTaufiqMethod(
|
70 |
+
model_path=os.path.join(model_path, "checkpoint-3"),
|
71 |
+
lowercase=True,
|
72 |
+
)
|
73 |
+
|
74 |
def run(text, source_style):
|
75 |
yield (
|
76 |
"(Memproses ...)",
|
|
|
82 |
|
83 |
start_time = time.time()
|
84 |
|
85 |
+
source_amr, *_ = t2a([text])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
source_amr.metadata = {}
|
87 |
source_amr_display = penman.encode(source_amr)
|
88 |
source_amr_display += f"\n\n({time.time() - start_time:.2f} s)"
|
|
|
95 |
)
|
96 |
|
97 |
triplets = sd.get_triplets(text)
|
|
|
|
|
|
|
|
|
98 |
triplets_display = "\n".join(f"({x[0]}, {x[1]}, {x[2]})" for x in triplets)
|
99 |
triplets_display += f"\n\n({time.time() - start_time:.2f} s)"
|
100 |
yield (
|
|
|
106 |
)
|
107 |
|
108 |
style_words = sd.get_style_words_from_triplets(triplets)
|
|
|
109 |
style_words_display = ", ".join(style_words)
|
110 |
style_words_display += f"\n\n({time.time() - start_time:.2f} s)"
|
111 |
yield (
|
|
|
127 |
"(Memproses ...)",
|
128 |
)
|
129 |
|
130 |
+
result, *_ = a2t([target_amr])
|
131 |
result += f"\n\n({time.time() - start_time:.2f} s)"
|
132 |
yield (
|
133 |
source_amr_display,
|