Enable Qwen reasoning by default

app.py

@@ -116,7 +116,7 @@ API_RESOURCES = {
             "prompt": "string",
             "max_new_tokens": "integer (1-256, default 64)",
             "temperature": "number (0-2, default 0.7)",
-            "thinking": "boolean (default false)",
+            "thinking": "boolean (default true; Qwen3.8 is a reasoning model)",
         },
         "output": "JSON containing generated text and CPU timing",
     },
@@ -601,9 +601,17 @@ def _generate_qwen38(prompt, max_new_tokens=64, temperature=0.7, thinking=False)
     processor, model = _load_qwen38()
     max_new_tokens = max(1, min(256, int(max_new_tokens)))
     temperature = max(0.0, min(2.0, float(temperature)))
-    content = prompt if thinking else prompt + "/no_think"
+    content = prompt
     messages = [{"role": "user", "content": content}]
-    text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+    try:
+        text = processor.apply_chat_template(
+            messages,
+            tokenize=False,
+            add_generation_prompt=True,
+            enable_thinking=True,
+        )
+    except TypeError:
+        text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
     inputs = processor(text=[text], return_tensors="pt")
     inputs = {key: value.to("cpu") if hasattr(value, "to") else value for key, value in inputs.items()}
     t0 = time.time()
@@ -617,7 +625,7 @@ def _generate_qwen38(prompt, max_new_tokens=64, temperature=0.7, thinking=False)
     elapsed = max(time.time() - t0, 0.001)
     input_tokens = int(inputs["input_ids"].shape[-1])
     generated = output[0, input_tokens:]
-    answer = processor.tokenizer.decode(generated, skip_special_tokens=True)
+    answer = processor.tokenizer.decode(generated, skip_special_tokens=True).strip()
     return {"text": answer, "prompt_tokens": input_tokens, "completion_tokens": int(generated.shape[-1]), "generation_tokens_per_sec": round(float(generated.shape[-1]) / elapsed, 2), "inference_time_sec": round(elapsed, 3), "device": "cpu", "dtype": "bfloat16", "thinking": bool(thinking)}
 
 
@@ -746,12 +754,11 @@ def _patched_create_app(blocks, **kwargs):
             return JSONResponse({"error": "Missing 'prompt'"}, status_code=400)
         job = _start_benchmark_job(
             "qwen38",
-            prompt,
-            False,
+            None,
             request={
-                "max_new_tokens": payload.get("max_new_tokens", 64),
+                "max_new_tokens": payload.get("max_new_tokens", 256),
                 "temperature": payload.get("temperature", 0.7),
-                "thinking": payload.get("thinking", False),
+                "thinking": payload.get("thinking", True),
             },
         )
         return JSONResponse({"status": "accepted", "job_id": job, "poll": "/respite/text/qwen38?job_id=" + job}, status_code=202)