coolpoodle

/

Qwen3-0.6B-Looped

Like 6

from example provided in the readme.md

python src/modules/llm.py 
Loading model...
Traceback (most recent call last):
  File "/home/uwu/Documents/programming/src/modules/llm.py", line 8, in <module>
    model = AutoModelForCausalLM.from_pretrained(
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/uwu/Documents/programming/.venv/lib/python3.12/site-packages/transformers/models/auto/auto_factory.py", line 340, in from_pretrained
    config, kwargs = AutoConfig.from_pretrained(
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/uwu/Documents/programming/.venv/lib/python3.12/site-packages/transformers/models/auto/configuration_auto.py", line 425, in from_pretrained
    config_class.register_for_auto_class()
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: type object 'Qwen3LoopConfig' has no attribute 'register_for_auto_class'

the code (the example provided in the readme of the project):

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "coolpoodle/Qwen3-0.6B-Looped"

print("Loading model...")
# trust_remote_code=True is essential for the custom architecture
model = AutoModelForCausalLM.from_pretrained(
    model_id, 
    trust_remote_code=True, 
    torch_dtype=torch.float16,
    device_map="auto"
)

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)

# Prompt
prompt = "The future of artificial intelligence is"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

# Generate
# use_cache=False is RECOMMENDED for Loop Attention to fully activate its mixing logic during generation
print("Generating...")
with torch.no_grad():
    outputs = model.generate(
        **inputs, 
        max_new_tokens=100, 
        do_sample=True, 
        temperature=0.7,
        use_cache=False 
    )

print("-" * 20)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
print("-" * 20)