[#1] fetch_alpha implemented

explore/explore_fetch_alpha.py

@@ -0,0 +1,10 @@
+from idiomify.fetchers import fetch_alpha
+
+
+def main():
+    model = fetch_alpha("overfit")
+    print(model.bart.config)
+
+
+if __name__ == '__main__':
+    main()

explore/explore_fetch_alpha_predict.py

@@ -0,0 +1,19 @@
+from transformers import BartTokenizer
+from builders import SourcesBuilder
+from fetchers import fetch_alpha
+
+
+def main():
+    model = fetch_alpha("overfit")
+    tokenizer = BartTokenizer.from_pretrained("facebook/bart-base")
+    lit2idi = [
+        ("my man", ""),
+        ("hello", "")
+    ]  # just some dummy stuff
+    srcs = SourcesBuilder(tokenizer)(lit2idi)
+    out = model.predict(srcs=srcs)
+    print(out)
+
+
+if __name__ == '__main__':
+    main()

explore/explore_fetch_epie.py

@@ -1,27 +0,0 @@
-
-from idiomify.fetchers import fetch_epie
-
-
-def main():
-    epie = fetch_epie()
-    idioms = set([
-        idiom
-        for idiom, _, _ in epie
-    ])
-
-    # so, what do you want? you want to build an idiom-masked language modeling?
-    for idiom, context, tag in epie:
-        print(idiom, context)
-
-    for idx, idiom in enumerate(idioms):
-        print(idx, idiom)
-
-    # isn't it better to just leave the idiom there, and have it guess what meaning it has?
-    # in that case, It may be better to use a generative model?
-    # but what would happen if you let it... just guess it?
-    # the problem with non-masking is that ... you give the model the answer.
-    # what you should rather do is... do something like...  find similar words.
-
-
-if __name__ == '__main__':
-    main()

idiomify/fetchers.py

@@ -5,7 +5,7 @@ import wandb
 import requests
 from typing import Tuple, List
 from wandb.sdk.wandb_run import Run
-from idiomify.paths import CONFIG_YAML, idioms_dir, literal2idiomatic
+from idiomify.paths import CONFIG_YAML, idioms_dir, literal2idiomatic, alpha_dir
 from idiomify.urls import (
     EPIE_IMMUTABLE_IDIOMS_URL,
     EPIE_IMMUTABLE_IDIOMS_CONTEXTS_URL,
@@ -15,9 +15,10 @@ from idiomify.urls import (
     EPIE_MUTABLE_IDIOMS_TAGS_URL,
     PIE_URL
 )
+from transformers import AutoModelForSeq2SeqLM, AutoConfig
+from models import Alpha
 
 
-# sources for dataset
 def fetch_epie(ver: str) -> List[Tuple[str, str, str]]:
     """
     It fetches the EPIE idioms, contexts, and tags from the web
@@ -85,6 +86,20 @@ def fetch_literal2idiomatic(ver: str, run: Run = None) -> List[Tuple[str, str]]:
         return [(row[0], row[1]) for row in reader]
 
 
+def fetch_alpha(ver: str, run: Run = None) -> Alpha:
+    if run:
+        artifact = run.use_artifact(f"alpha:{ver}", type="model")
+    else:
+        artifact = wandb.Api().artifact(f"eubinecto/idiomify/alpha:{ver}", type="model")
+    config = artifact.metadata
+    artifact_dir = artifact.download(root=alpha_dir(ver))
+    ckpt_path = path.join(artifact_dir, "model.ckpt")
+    bart = AutoModelForSeq2SeqLM.from_config(AutoConfig.from_pretrained(config['bart']))
+    with open(ckpt_path, 'r') as fh:
+        alpha = Alpha.load_from_checkpoint(ckpt_path, bart=bart)
+    return alpha
+
+
 def fetch_config() -> dict:
     with open(str(CONFIG_YAML), 'r', encoding="utf-8") as fh:
         return yaml.safe_load(fh)

idiomify/paths.py

@@ -15,7 +15,3 @@ def literal2idiomatic(ver: str) -> Path:
 
 def alpha_dir(ver: str) -> Path:
     return ARTIFACTS_DIR / f"alpha_{ver}"
-
-
-def gamma_dir(ver: str) -> Path:
-    return ARTIFACTS_DIR / f"beta_{ver}"