[CI] Refactor import paths in dequantization examples to use dequantize_utils (#914)

* Update requirements and refactor benchmark script for deepseek_nsa example

- Updated the requirements.txt to specify a fixed commit for the flash-linear-attention repository.
- Refactored import paths in benchmark_nsa_fwd.py for better organization.
- Added a new function to generate configurations for autotuning.
- Modified the tilelang_sparse_attention function to accept parameters for block size, number of stages, and threads, enhancing flexibility.
- Changed allocation of shared memory for accumulators to optimize performance.

* Refactor import paths in dequantization examples to use dequantize_utils

- Updated import statements in multiple dequantization example scripts to replace references to the removed utils.py file with the new dequantize_utils module.
- Ensured consistency across example scripts for better organization and maintainability.

examples/dequantize_gemm/example_dequant_gemm_bf16_fp4_hopper.py

@@ -4,7 +4,7 @@ from tilelang import tvm as tvm
 from tvm import DataType
 from tvm import tir
 import torch
-from utils import torch_convert_bit_twiddling, torch_convert
+from dequantize_utils import torch_convert_bit_twiddling, torch_convert
 
 
 def get_configs():

examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_hopper.py

@@ -4,7 +4,7 @@ from tilelang import tvm as tvm
 from tvm import DataType
 from tvm import tir
 import torch
-from utils import torch_convert_bit_twiddling, torch_convert
+from dequantize_utils import torch_convert_bit_twiddling, torch_convert
 
 
 def _tir_u8_to_f4_to_bf16(nbit: int, val: tir.PrimExpr, pos: tir.PrimExpr, scale: tir.PrimExpr,

examples/dequantize_gemm/example_dequant_gemm_bf16_mxfp4_hopper_tma.py

@@ -4,7 +4,7 @@ from tilelang import tvm as tvm
 from tvm import DataType
 from tvm import tir
 import torch
-from utils import torch_convert_bit_twiddling, torch_convert
+from dequantize_utils import torch_convert_bit_twiddling, torch_convert
 
 
 def _tir_u8_to_f4_to_bf16(nbit: int, val: tir.PrimExpr, pos: tir.PrimExpr, scale: tir.PrimExpr,

examples/dequantize_gemm/example_dequant_groupedgemm_bf16_mxfp4_hopper.py

@@ -4,7 +4,7 @@ from tilelang.quantize import _tir_u8_to_f4_to_bf16
 from tilelang import tvm as tvm
 from tvm import DataType
 import torch
-from utils import torch_convert_bit_twiddling, assert_similar
+from dequantize_utils import torch_convert_bit_twiddling, assert_similar
 from tilelang.autotuner import set_autotune_inputs