feat: add linear attention masking support for Qwen3.5 models

[dogwood-flo]

Summary

• Add a function to Blockwise.py that applies masking compatible with linear attention.
• Detect Qwen3.5 attention modules and apply the appropriate masking logic.

Validation

Verified the implementation with the following script:

import torch

from onecomp import (
    GPTQ,
    CalibrationConfig,
    ModelConfig,
    Runner,
    load_quantized_model,
    setup_logger,
)

setup_logger()

MODEL_ID = "Qwen/Qwen3.6-27B"
SAVE_DIR = "./Qwen3.6-27B_gptq4"

model_config = ModelConfig(model_id=MODEL_ID, device="cuda:0")

gptq = GPTQ(
    wbits=4,
    groupsize=128,
)

runner = Runner(
    model_config=model_config,
    quantizer=gptq,
    qep=True,
    calibration_config=CalibrationConfig(max_length=512, num_calibration_samples=128),
)

runner.run()

runner.save_quantized_model(SAVE_DIR)
print(f"\nQuantized model saved to: {SAVE_DIR}")

model, tokenizer = load_quantized_model(SAVE_DIR, torch_dtype=torch.bfloat16)
print(f"Loaded model type : {type(model).__name__}")
print(f"Loaded model device: {next(model.parameters()).device}")

prompt = "Fujitsu is"
inputs = tokenizer(prompt, return_tensors="pt").to(next(model.parameters()).device)

with torch.no_grad():
    output_ids = model.generate(**inputs, max_new_tokens=32, do_sample=False)

generated = tokenizer.decode(output_ids[0], skip_special_tokens=True)
print(f"\nPrompt   : {prompt}")
print(f"Generated: {generated}")