import torch
from unsloth import FastLanguageModel
from transformers import TextStreamer

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "ubermenchh/Qwen2.5-3B-open-r1-math",
    max_seq_length = 1024,
    dtype = torch.bfloat16,
    load_in_4bit = True,
)
FastLanguageModel.for_inference(model) 

SYSTEM_PROMPT = """
Respond in the following format:
<think>
...
</think>
<answer>
...
</answer>
"""

test_question = """
Let $z \in \mathbf{C}$, satisfying the condition $a z^{n}+b \mathrm{i} z^{n-1}+b \mathrm{i} z-a=0, a, b \in \mathbf{R}, m \in$ $\mathbf{N}$, find $|z|$.
"""

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": test_question},
]
input_ids = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt = True,
    return_tensors = "pt",
).to("cuda")

text_streamer = TextStreamer(tokenizer, skip_prompt = True)
_ = model.generate(input_ids, streamer = text_streamer, max_new_tokens = 2048, pad_token_id = tokenizer.eos_token_id)

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained(
  "ubermenchh/Qwen2.5-3B-open-r1-math",
  torch_dtype=torch.bfloat16,
  device_map="auto",
  trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained(
  "ubermenchh/Qwen2.5-3B-open-r1-math",
  trust_remote_code=True
)

SYSTEM_PROMPT = """
Respond in the following format:
<think>
...
</think>
<answer>
...
</answer>
"""

problem = "Let $z \in \mathbf{C}$, satisfying the condition $a z^{n}+b \mathrm{i} z^{n-1}+b \mathrm{i} z-a=0, a, b \in \mathbf{R}, m \in$ $\mathbf{N}$, find $|z|$."
prompt = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": problem}
]

input_text = tokenizer.apply_chat_template(prompt, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(input_text, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs, 
    max_new_tokens=3000, 
    temperature=1.3, 
    num_return_sequences=1, 
    pad_token_id=tokenizer.pad_token_id,
    eos_token_id=tokenizer.eos_token_id
)

response = tokenizer.decode(outputs[0], skip_special_tokens=True)

print("Question:\n", problem)
print("\n\nResponse:\n", response)