from transformers import pipeline
model_id = "tiny-random/qwen3-tp"
pipe = pipeline(
    "text-generation", model=model_id, device="cuda",
    trust_remote_code=True, max_new_tokens=3,
)
print(pipe("Hello World!"))


from transformers import AutoModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto"
)
prompt = "Give me a short introduction to large language model."
messages = [
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=True  # Switches between thinking and non-thinking modes. Default is True.
)
print(text)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=128
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()
try:
    # rindex finding 151668 (</think>)
    index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
    index = 0
thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")
print("thinking content:", thinking_content)
print("content:", content)

Codes to create this repo:

python

import torch

from transformers import (
    AutoConfig,
    AutoModelForCausalLM,
    AutoTokenizer,
    GenerationConfig,
    pipeline,
    set_seed,
)

source_model_id = "Qwen/Qwen3-32B"
save_folder = "/tmp/tiny-random/qwen3-tp"

tokenizer = AutoTokenizer.from_pretrained(
    source_model_id, trust_remote_code=True,
)
tokenizer.save_pretrained(save_folder)

config = AutoConfig.from_pretrained(
    source_model_id, trust_remote_code=True,
)
config._name_or_path = source_model_id
config.hidden_size = 8
config.intermediate_size = 32
config.head_dim = 32
config.num_key_value_heads = 4
config.num_attention_heads = 8
config.num_hidden_layers = 2
config.max_window_layers = 1
config.tie_word_embeddings = False
config.layer_types = ['full_attention'] * 2
model = AutoModelForCausalLM.from_config(
    config,
    torch_dtype=torch.bfloat16,
    trust_remote_code=True,
)
model.generation_config = GenerationConfig.from_pretrained(
    source_model_id, trust_remote_code=True,
)
set_seed(42)
with torch.no_grad():
    for name, p in sorted(model.named_parameters()):
        torch.nn.init.normal_(p, 0, 0.1)
        print(name, p.shape)
model.save_pretrained(save_folder)