__

app.py

@@ -3,11 +3,24 @@ from transformers import (
     AutoTokenizer, AutoModelForCausalLM)
 st.balloons()
 
-@st.cache
-def f():
-    return 1
+device = 'cuda'
 
-st.text(f())
+@st.cache
+def model_init():
+    tokenizer = AutoTokenizer.from_pretrained(
+        "ckip-joint/bloom-1b1-zh")
+    model = AutoModelForCausalLM.from_pretrained(
+        "ckip-joint/bloom-1b1-zh",
+        # Ref.: Eric, Thanks!
+        # torch_dtype="auto", 
+        # device_map="auto",
+    # Ref. for `half`: Chan-Jan, Thanks!
+    ).eval().to(device)
+    return tok, model
+
+tok, model = model_init()
+
+st.text(str(repr(model)))
 
 # DONE 6.1s