outlines with gemma4 output wrong answer

!python3 -m pip install -U transformers bitsandbytes outlines
import huggingface_hub, transformers, torch, outlines, builtins
huggingface_hub.snapshot_download(repo_id='exact-railcar/gemma', local_dir='/tmp/kaggle')
processor = transformers.Gemma4Processor.from_pretrained('/tmp/kaggle')
model = transformers.Gemma4ForConditionalGeneration.from_pretrained('/tmp/kaggle', device_map='auto', dtype=torch.bfloat16)
messages = [
    {"role": "user", "content": "What is 1 + 1"},
]

# Process input
inputs = processor.apply_chat_template(
    messages,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    add_generation_prompt=True,
    enable_thinking=True
).to(model.device)
input_len = inputs["input_ids"].shape[-1]

# Generate output
outputs = model.generate(**inputs, max_new_tokens=1024)
response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)

# Parse output
processor.parse_response(response, prefix=inputs["input_ids"])

{'role': 'assistant',
 'thinking': 'The user is asking for the sum of 1 and 1.\nBasic arithmetic. $1 + 1 = 2$.',
 'content': '1 + 1 = 2'}

model = outlines.from_transformers(transformers.Gemma4ForConditionalGeneration.from_pretrained('/tmp/kaggle', device_map='auto', dtype=torch.bfloat16), transformers.Gemma4Processor.from_pretrained('/tmp/kaggle'))
model(outlines.inputs.Chat([{"role": "user", "content": "What is 1 + 1"}]), output_type=builtins.int, temperature=0.1, max_new_tokens=1024)

see above

just 2