[Cleanup] Remove `tilelang.disable_cache()` calls from examples and tests (#1088)

* [Cleanup] Remove `tilelang.disable_cache()` calls from example scripts

* lint

* lint

examples/blocksparse_attention/example_tilelang_sparse_gqa_decode_paged.py

@@ -11,8 +11,6 @@ import math
 
 from heuristic import num_splits_heuristic
 
-tilelang.disable_cache()
-
 
 def flashattn(batch, heads, heads_kv, dim, dim_v):
     scale = (1.0 / dim)**0.5 * 1.44269504  # log2(e)

examples/cast/example_per_token_cast_to_fp8.py

@@ -4,8 +4,6 @@ import tilelang.language as T
 from typing import Tuple
 from tilelang.utils.tensor import torch_assert_close
 
-tilelang.disable_cache()
-
 
 @tilelang.jit(out_idx=[1, 2])
 def per_token_cast_to_fp8(M, N, blk_m):

examples/deepseek_mla/amd/benchmark_mla_decode_amd_tilelang.py

@@ -5,8 +5,6 @@ import tilelang.language as T
 from einops import rearrange, einsum
 import argparse
 
-tilelang.disable_cache()
-
 
 def get_configs():
     import itertools

examples/elementwise/example_elementwise_add.py

@@ -5,8 +5,6 @@ import tilelang
 import tilelang.language as T
 from tilelang.autotuner import AutoTuner
 
-tilelang.disable_cache()
-
 
 def ref_program(x, y):
     return x + y

examples/flash_attention/example_gqa_bwd_tma_reduce_varlen.py

@@ -7,7 +7,6 @@ import argparse
 from einops import rearrange, repeat
 from bert_padding import pad_input, unpad_input
 
-# tilelang.disable_cache()
 torch.manual_seed(1)
 
 

examples/gdn/example_chunk_delta_bwd.py

@@ -24,8 +24,6 @@ import torch.nn.functional as F
 torch.random.manual_seed(0)
 # torch.set_printoptions(profile="full")
 
-tilelang.disable_cache()
-
 from utils import *
 
 

examples/gdn/example_chunk_delta_h.py

@@ -32,8 +32,6 @@ from utils import *
 
 torch.random.manual_seed(0)
 
-tilelang.disable_cache()
-
 
 def prepare_input(
     B,

examples/gdn/example_chunk_o.py

@@ -19,8 +19,6 @@ import torch
 
 torch.random.manual_seed(1)
 
-tilelang.disable_cache()
-
 
 def prepare_input(
     B,

examples/gdn/example_chunk_o_bwd.py

@@ -26,8 +26,6 @@ from utils import *
 torch.random.manual_seed(0)
 # torch.set_printoptions(profile="full")
 
-tilelang.disable_cache()
-
 
 def prepare_input_fake(
     B,

examples/gdn/example_chunk_scaled_dot_kkt.py

@@ -20,8 +20,6 @@ import torch
 torch.set_printoptions(profile="full")
 torch.random.manual_seed(0)
 
-tilelang.disable_cache()
-
 
 def prepare_input(
     B,

examples/gdn/example_cumsum.py

@@ -18,8 +18,6 @@ except ImportError:
 
 import torch
 
-tilelang.disable_cache()
-
 
 @tilelang.jit(
     out_idx=[-1],

examples/gdn/example_wy_fast.py

@@ -19,8 +19,6 @@ import torch
 
 torch.random.manual_seed(1)
 
-tilelang.disable_cache()
-
 
 def prepare_input(B, S, H, DK, DV, chunk_size, input_dtype, output_dtype, gate_dtype=torch.float32):
     BS = chunk_size

examples/gdn/example_wy_fast_bwd_split.py

@@ -22,8 +22,6 @@ import torch.nn.functional as F
 torch.random.manual_seed(0)
 torch.set_printoptions(profile="full")
 
-tilelang.disable_cache()
-
 
 def prepare_input_fake(
     B,

examples/gemm_sm100/gemm_tcgen5mma.py

@@ -2,8 +2,6 @@ import torch
 import tilelang
 import tilelang.language as T
 
-tilelang.disable_cache()
-
 
 def matmul(
     M,

examples/grouped_gemm/example_grouped_gemm_bwd.py

@@ -4,8 +4,6 @@ import argparse
 import tilelang
 import tilelang.language as T
 
-tilelang.disable_cache()
-
 
 @tilelang.jit(
     out_idx=[2], pass_configs={

examples/minference/example_vertical_slash_sparse_attn.py

@@ -12,8 +12,6 @@ import tilelang
 import tilelang.language as T
 from tilelang.profiler import do_bench
 
-tilelang.disable_cache()
-
 
 @tilelang.jit(out_idx=[3])
 def _tl_vs_sparse_flashattn(batch, heads, seq_len, dim, vertical_size, slash_size):

examples/warp_specialize/example_warp_specialize_flashmla.py

@@ -6,8 +6,6 @@ import tilelang.language as T
 from einops import rearrange, einsum
 import argparse
 
-tilelang.disable_cache()
-
 
 @tilelang.jit(out_idx=[6])
 def flashattn(batch, heads, kv_head_num, seqlen_kv, dim, pe_dim, block_N, block_H, num_split):

examples/warp_specialize/example_warp_specialize_gemm_copy_gemm_0_1.py

 import tilelang
 import tilelang.language as T
 
-tilelang.disable_cache()
-
 
 # add decorator @tilelang.jit if you want to return a torch function
 # @tilelang.jit

examples/warp_specialize/example_warp_specialize_gemm_softpipe_stage2.py

@@ -50,8 +50,6 @@ def main(M=16384, N=16384, K=16384):
 
     jit_kernel = matmul(M, N, K, block_M, block_N, block_K)
 
-    tilelang.disable_cache()
-
     # 3. Test the kernel in Python with PyTorch data
     import torch
 

maint/precision/compare_ops.py

@@ -17,8 +17,6 @@ from triton.language.extra import libdevice
 import tilelang
 import tilelang.language as T
 
-tilelang.disable_cache()
-
 from tilelang.contrib import nvcc
 from tilelang.utils.target import determine_target