add qwen3's

app.py

@@ -29,6 +29,12 @@ arg_dict = {
         "meta-llama/Llama-3.1-8B",
         "meta-llama/Llama-3.2-3B",
         "meta-llama/Llama-3.2-1B",
+        "Qwen/Qwen3-8B",
+        "Qwen/Qwen3-4B",
+        "Qwen/Qwen3-1.7B",
+        "Qwen/Qwen3-0.6B",
+        "Qwen/Qwen3-4B-Instruct-2507",
+        "Qwen/Qwen3-4B-Thinking-2507",
     ],
     # 'load_fp16' : True,
     'load_fp16' : False,

demo_watermark.py

@@ -201,7 +201,7 @@ def load_model(args):
     """Load and return the model and tokenizer"""
 
     args.is_seq2seq_model = any([(model_type in args.model_name_or_path.lower()) for model_type in ["t5","T0"]])
-    args.is_decoder_only_model = any([(model_type in args.model_name_or_path.lower()) for model_type in ["gpt","opt","bloom","llama"]])
+    args.is_decoder_only_model = any([(model_type in args.model_name_or_path.lower()) for model_type in ["gpt","opt","bloom","llama","qwen"]])
     if args.is_seq2seq_model:
         model = AutoModelForSeq2SeqLM.from_pretrained(args.model_name_or_path)
     elif args.is_decoder_only_model:
@@ -778,8 +778,10 @@ def run_gradio(args, model=None, device=None, tokenizer=None):
             # else:
             return AutoTokenizer.from_pretrained(model_name_or_path)
         
-        def update_model(session_state): 
-            model, _, _ = load_model(session_state)
+        def update_model(state, old_model): 
+            del old_model 
+            torch.cuda.empty_cache()
+            model, _, _ = load_model(state)
             return model
         
         def check_model(value): return value if (value!="" and value is not None) else args.model_name_or_path
@@ -802,7 +804,7 @@ def run_gradio(args, model=None, device=None, tokenizer=None):
         ).then(
             update_tokenizer,inputs=[model_selector], outputs=[session_tokenizer]
         ).then(
-            update_model,inputs=[session_args], outputs=[session_model]
+            update_model,inputs=[session_args, session_model], outputs=[session_model]
         ).then(
             lambda value: str(value), inputs=[session_args], outputs=[current_parameters]
         )