feat: integrate StyleDetector, disable TextToAMR for fast test

app.py

@@ -1,15 +1,18 @@
+import gdown
 from git import Repo
 import gradio as gr
 from huggingface_hub import snapshot_download
+import os
 import penman
 import sys
 
-Repo.clone_from("https://github.com/AbdiHaryadi/amr-tst-indo.git", "amr-tst-indo")
+if not os.path.exists("amr-tst-indo"):
+    Repo.clone_from("https://github.com/AbdiHaryadi/amr-tst-indo.git", "amr-tst-indo")
 sys.path.append("./amr-tst-indo")
 
 from text_to_amr import TextToAMR
+from style_detector import StyleDetector
 
-# Gold v2
 amr_parsing_model_name = "mbart-en-id-smaller-indo-amr-parsing-translated-nafkhan"
 snapshot_download(
     repo_id=f"abdiharyadi/{amr_parsing_model_name}",
@@ -21,20 +24,36 @@ snapshot_download(
 )
 t2a = TextToAMR(model_name=amr_parsing_model_name)
 
+gdown.download(
+    "https://drive.google.com/uc?id=1J_6PbYsQ6Kl4Qfs1wBVwd52_r9uTpIxx",
+    "./model-best.pt"
+)
+sd = StyleDetector(
+    config_path="./amr-tst-indo/indonesian-aste-generative/resources/exp-v2/exp-m0.yaml",
+    model_path="./model-best.pt"
+)
+
 def run(text, source_style):
-    source_amr, *_ = t2a([text])
-    source_amr.metadata = {}
-    source_amr_display = penman.encode(source_amr)
+    # source_amr, *_ = t2a([text])
+    # source_amr.metadata = {}
+    # source_amr_display = penman.encode(source_amr)
+    source_amr_display = "(z0 / halo)"
     yield source_amr_display, "...", "...", "...", "..."
 
-    triplets = [
-        ("kamar", "sangat bagus", "positif"),
-        ("kamar", "bersih", "positif")
-    ]
-    triplets_display = "\n".join(f"({x[0]}, {x[1]}, {x[2]})" for x in triplets)
-    yield source_amr_display, triplets_display, "...", "...", "..."
+    triplet_display_dict = {"data": "..."}
+    def triplets_callback(triplets: list):
+        triplet_display_dict["data"] = "\n".join(f"({x[0]}, {x[1]}, {x[2]})" for x in triplets)
+        triplets_display = triplet_display_dict["data"]
+        yield source_amr_display, triplets_display, "...", "...", "..."
 
-    style_words = ["bagus", "bersih"]
+    style_words = sd(text, triplets_callback=triplets_callback)
+    # style_words = ["bagus", "bersih"]
+    # triplets = [
+    #     ("kamar", "sangat bagus", "positif"),
+    #     ("kamar", "bersih", "positif")
+    # ]
+    # triplets_display = "\n".join(f"({x[0]}, {x[1]}, {x[2]})" for x in triplets)
+    triplets_display = triplet_display_dict["data"]
     style_words_display = ", ".join(style_words)
     yield source_amr_display, triplets_display, style_words_display, "...", "..."
 

requirements.txt

@@ -1,6 +1,7 @@
 gitpython
 scipy>=1.7.2
 scikit_learn>=1.0.2
+gdown
 
 accelerate==0.32.1
 amrlib==0.8.0