🔐 add auth token and experimental models

Signed-off-by: peter szemraj <peterszemraj@gmail.com>

app.py

@@ -64,8 +64,9 @@ nltk.download("popular", force=True, quiet=True)
 MODEL_OPTIONS = [
     "pszemraj/long-t5-tglobal-base-16384-book-summary",
     "pszemraj/long-t5-tglobal-base-sci-simplify",
-    "pszemraj/long-t5-tglobal-base-sci-simplify-elife",
-    "pszemraj/long-t5-tglobal-base-16384-booksci-summary-v1",
+    "pszemraj/long-t5-tglobal-base-summary-souffle-16384-loD",
+    "pszemraj/long-t5-tglobal-base-summary-souffle-16384-neftune_0.3",
+    "pszemraj/long-t5-tglobal-base-summary-souffle-16384-neftune_0.6",
     "pszemraj/pegasus-x-large-book-summary",
 ]  # models users can choose from
 BEAM_OPTIONS = [2, 3, 4]  # beam sizes users can choose from

summarize.py

@@ -2,6 +2,7 @@
 summarize - a module for summarizing text using a model from the Hugging Face model hub
 """
 import logging
+import os
 import pprint as pp
 
 logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(message)s")
@@ -23,10 +24,14 @@ def load_model_and_tokenizer(model_name: str) -> tuple:
     device = "cuda" if torch.cuda.is_available() else "cpu"
     model = AutoModelForSeq2SeqLM.from_pretrained(
         model_name,
+        use_auth_token=os.environ.get("HF_TOKEN", None),
     ).to(device)
     model = model.eval()
 
-    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    tokenizer = AutoTokenizer.from_pretrained(
+        model_name,
+        use_auth_token=os.environ.get("HF_TOKEN", None),
+    )
 
     logging.info(f"Loaded model {model_name} to {device}")