Revert FP16 autocast — increases TTFA on first request

autocast triggers fp16 kernel selection at first call for each tensor
shape. Since the warmup uses short text, real requests re-trigger
selection and are slower net. Keeping FP32 + conditionals cache.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

engine.py

@@ -116,8 +116,7 @@ def synthesize(
         kwargs["cfg_weight"] = cfg_weight
 
     with torch.inference_mode():
-        with torch.amp.autocast(device_type="cuda", dtype=torch.float16):
+        wav = chatterbox_model.generate(text=text, **kwargs)
-            wav = chatterbox_model.generate(text=text, **kwargs)
 
     if torch.cuda.is_available():
         torch.cuda.synchronize()