deepseek-ai/TileKernels
File Size
The distribution of size of files (measured in lines of code).
File Size Overall
  • There are 140 files with 16,231 lines of code.
    • 0 very long files (0 lines of code)
    • 1 long files (656 lines of code)
    • 23 medium size files (6,679 lines of code)
    • 40 small files (5,552 lines of code)
    • 76 very small files (3,344 lines of code)
0% | 4% | 41% | 34% | 20%
Legend:
1001+
501-1000
201-500
101-200
1-100


explore: grouped by folders | grouped by size | sunburst | 3D view
File Size per Extension
1001+
501-1000
201-500
101-200
1-100
py0% | 4% | 41% | 34% | 20%
yaml0% | 0% | 0% | 0% | 100%
File Size per Logical Decomposition
primary
1001+
501-1000
201-500
101-200
1-100
tile_kernels0% | 4% | 41% | 34% | 20%
ROOT0% | 0% | 0% | 0% | 100%
Longest Files (Top 50)
File# lines# units
benchmark.py
in tile_kernels/testing/pytest
 656 45
swiglu_backward_asc.py
in tile_kernels/quant
 445 8
pre_split_mixes_asc.py
in tile_kernels/mhc
 414 5
norm_forward_asc.py
in tile_kernels/quant
 379 11
norm_fn_asc.py
in tile_kernels/mhc
 369 5
pre_big_fuse_asc.py
in tile_kernels/mhc
 354 4
swiglu_forward_asc.py
in tile_kernels/quant
 347 12
per_token_cast_asc.py
in tile_kernels/quant
 336 11
330 17
gpu_mem.py
in tile_kernels/testing/pytest
 319 21
engram_gate_bwd_cuda.py
in tile_kernels/engram
 274 8
moe_topk_gate_forward_asc.py
in tile_kernels/moe
 265 2
per_channel_cast_asc.py
in tile_kernels/quant
 262 5
common.py
in tile_kernels/quant
 261 24
swiglu_forward_cuda.py
in tile_kernels/quant
 260 2
cast.py
in tile_kernels/torch
 255 8
engram_sinkhorn_asc.py
in tile_kernels/engram
 252 8
sinkhorn_asc.py
in tile_kernels/mhc
 245 5
engram_gate_fwd_asc.py
in tile_kernels/engram
 233 2
220 3
per_block_cast_lossless_asc.py
in tile_kernels/quant
 218 11
norm_forward_cuda.py
in tile_kernels/quant
 217 3
engram_sinkhorn_cuda.py
in tile_kernels/engram
 213 11
swiglu_backward_cuda.py
in tile_kernels/quant
 211 2
per_block_cast_asc.py
in tile_kernels/quant
 195 8
norm_forward_kernel.py
in tile_kernels/quant
 194 4
norm_backward_cuda.py
in tile_kernels/quant
 192 3
cast_back_asc.py
in tile_kernels/quant
 190 6
rope_asc.py
in tile_kernels/transform
 185 5
engram_gate_bwd_asc.py
in tile_kernels/engram
 181 2
swiglu_backward_kernel.py
in tile_kernels/quant
 180 3
per_token_cast_cuda.py
in tile_kernels/quant
 179 4
pre_split_mixes_cuda.py
in tile_kernels/mhc
 177 3
norm_backward_asc.py
in tile_kernels/quant
 172 5
post_asc.py
in tile_kernels/mhc
 164 2
engram_gate_fwd_cuda.py
in tile_kernels/engram
 163 4
per_token_cast_kernel.py
in tile_kernels/quant
 149 4
148 1
swiglu.py
in tile_kernels/torch
 143 5
topk.py
in tile_kernels/torch
 141 4
multilayer_recompute_asc.py
in tile_kernels/mhc
 138 1
per_channel_cast_cuda.py
in tile_kernels/quant
 136 2
per_block_cast_kernel.py
in tile_kernels/quant
 134 4
generator.py
in tile_kernels/testing
 131 11
norm_fn_cuda.py
in tile_kernels/mhc
 131 4
sinkhorn_cuda.py
in tile_kernels/mhc
 130 2
post_cuda.py
in tile_kernels/mhc
 127 2
per_block_cast_cuda.py
in tile_kernels/quant
 122 4
swiglu_forward_kernel.py
in tile_kernels/quant
 120 3
norm_fn_kernel.py
in tile_kernels/mhc
 117 4
Large Files That Change Often (Top 20)
To change a file, an AI coding agent (or a person) has to read it: these files are read the most for the changes in the past year (lines × changes).
File# lines≈ tokens
to read		# changes
(1y)		≈ lines edited
per change		lines read per
line changed
benchmark.py
in tile_kernels/testing/pytest
 656 ~3k–10k 2 ~1,000 ~1
swiglu_backward_asc.py
in tile_kernels/quant
 445 ~3k–7k 2 ~500 ~1
pre_split_mixes_asc.py
in tile_kernels/mhc
 414 ~2k–7k 2 ~500 ~1
norm_forward_asc.py
in tile_kernels/quant
 379 ~2k–7k 2 ~500 ~1
norm_fn_asc.py
in tile_kernels/mhc
 369 ~2k–7k 2 ~500 ~1
pre_big_fuse_asc.py
in tile_kernels/mhc
 354 ~2k–5k 2 ~500 ~1
swiglu_forward_asc.py
in tile_kernels/quant
 347 ~2k–5k 2 ~300 ~1
per_token_cast_asc.py
in tile_kernels/quant
 336 ~2k–5k 2 ~300 ~1
330 ~2k–5k 2 ~300 ~1
gpu_mem.py
in tile_kernels/testing/pytest
 319 ~2k–5k 2 ~500 ~1
engram_gate_bwd_cuda.py
in tile_kernels/engram
 274 ~1.5k–5k 2 ~300 ~1
moe_topk_gate_forward_asc.py
in tile_kernels/moe
 265 ~1.5k–5k 2 ~300 ~1
per_channel_cast_asc.py
in tile_kernels/quant
 262 ~1.5k–5k 2 ~300 ~1
common.py
in tile_kernels/quant
 261 ~1.5k–5k 2 ~300 ~1
swiglu_forward_cuda.py
in tile_kernels/quant
 260 ~1.5k–5k 2 ~300 ~1
cast.py
in tile_kernels/torch
 255 ~1.5k–5k 2 ~150 ~2
engram_sinkhorn_asc.py
in tile_kernels/engram
 252 ~1.5k–5k 2 ~300 ~1
sinkhorn_asc.py
in tile_kernels/mhc
 245 ~1.5k–5k 2 ~300 ~1
engram_gate_fwd_asc.py
in tile_kernels/engram
 233 ~1.5k–5k 2 ~300 ~1
220 ~1.5k–5k 2 ~300 ~1

Tokens are lines × 7 to 14 (analysis.fileReadsForChanges). Lines edited per change are the file's lines added and deleted divided by its commits over the whole history. Agents read about 2,000 lines at a time: when a 17,000-line file was split into modules, the tokens an agent read for the same change fell by 83% (The Economic Benefit of Refactoring, 2026).

Files With Most Units (Top 50)
File# lines# units
benchmark.py
in tile_kernels/testing/pytest
 656 45
common.py
in tile_kernels/quant
 261 24
gpu_mem.py
in tile_kernels/testing/pytest
 319 21
config.py
in tile_kernels
 90 18
330 17
swiglu_forward_asc.py
in tile_kernels/quant
 347 12
norm_forward_asc.py
in tile_kernels/quant
 379 11
per_token_cast_asc.py
in tile_kernels/quant
 336 11
per_block_cast_lossless_asc.py
in tile_kernels/quant
 218 11
engram_sinkhorn_cuda.py
in tile_kernels/engram
 213 11
generator.py
in tile_kernels/testing
 131 11
sqrt_softplus.py
in tile_kernels/moe
 106 10
bench.py
in tile_kernels/testing
 84 9
per_block_cast_asc.py
in tile_kernels/quant
 195 8
swiglu_backward_asc.py
in tile_kernels/quant
 445 8
engram_sinkhorn_asc.py
in tile_kernels/engram
 252 8
engram_gate_bwd_cuda.py
in tile_kernels/engram
 274 8
cast.py
in tile_kernels/torch
 255 8
mhc.py
in tile_kernels/torch
 86 8
cast_asc.py
in tile_kernels/quant
 107 7
cast_back_asc.py
in tile_kernels/quant
 190 6
engram.py
in tile_kernels/torch
 108 6
xdist_failfast.py
in tile_kernels/testing/pytest
 59 6
utils.py
in tile_kernels
 24 6
per_channel_cast_asc.py
in tile_kernels/quant
 262 5
rstd.py
in tile_kernels/quant
 65 5
norm_backward_asc.py
in tile_kernels/quant
 172 5
swiglu.py
in tile_kernels/torch
 143 5
pre_split_mixes_asc.py
in tile_kernels/mhc
 414 5
sinkhorn_asc.py
in tile_kernels/mhc
 245 5
norm_fn_asc.py
in tile_kernels/mhc
 369 5
rope_asc.py
in tile_kernels/transform
 185 5
per_block_cast_cuda.py
in tile_kernels/quant
 122 4
per_block_cast_lossless_cuda.py
in tile_kernels/quant
 108 4
norm_forward_kernel.py
in tile_kernels/quant
 194 4
per_token_cast_kernel.py
in tile_kernels/quant
 149 4
per_token_cast_cuda.py
in tile_kernels/quant
 179 4
per_block_cast_kernel.py
in tile_kernels/quant
 134 4
engram_gate_fwd_cuda.py
in tile_kernels/engram
 163 4
engram_sinkhorn_kernel.py
in tile_kernels/engram
 112 4
topk.py
in tile_kernels/torch
 141 4
common.py
in tile_kernels/testing/pytest
 25 4
numeric.py
in tile_kernels/testing
 52 4
expand_kernel.py
in tile_kernels/mhc
 36 4
norm_fn_kernel.py
in tile_kernels/mhc
 117 4
norm_fn_cuda.py
in tile_kernels/mhc
 131 4
pre_big_fuse_asc.py
in tile_kernels/mhc
 354 4
swiglu_backward_kernel.py
in tile_kernels/quant
 180 3
swiglu_forward_kernel.py
in tile_kernels/quant
 120 3
norm_backward_cuda.py
in tile_kernels/quant
 192 3
Files With Long Lines (Top 50)

There are 78 files with lines longer than 120 characters. In total, there are 351 long lines.

File# lines# units# long lines
swiglu_backward_asc.py
in tile_kernels/quant
 445 8 27
pre_split_mixes_asc.py
in tile_kernels/mhc
 414 5 18
norm_backward_cuda.py
in tile_kernels/quant
 192 3 16
per_block_cast_lossless_cuda.py
in tile_kernels/quant
 108 4 12
per_token_cast_asc.py
in tile_kernels/quant
 336 11 12
swiglu_forward_asc.py
in tile_kernels/quant
 347 12 12
per_channel_cast_asc.py
in tile_kernels/quant
 262 5 11
gpu_mem.py
in tile_kernels/testing/pytest
 319 21 11
norm_fn_asc.py
in tile_kernels/mhc
 369 5 11
norm_forward_kernel.py
in tile_kernels/quant
 194 4 10
engram_gate_bwd_cuda.py
in tile_kernels/engram
 274 8 10
330 17 9
per_token_cast_cuda.py
in tile_kernels/quant
 179 4 9
cast.py
in tile_kernels/torch
 255 8 9
moe_topk_gate_forward_asc.py
in tile_kernels/moe
 265 2 9
swiglu_backward_cuda.py
in tile_kernels/quant
 211 2 8
norm_forward_cuda.py
in tile_kernels/quant
 217 3 8
norm_backward_kernel.py
in tile_kernels/quant
 62 1 8
multilayer_recompute_asc.py
in tile_kernels/mhc
 138 1 7
topk_gate_asc.py
in tile_kernels/moe
 117 1 7
per_block_cast_cuda.py
in tile_kernels/quant
 122 4 6
cast_back_asc.py
in tile_kernels/quant
 190 6 6
110 2 6
per_block_cast_asc.py
in tile_kernels/quant
 195 8 5
per_block_cast_lossless_asc.py
in tile_kernels/quant
 218 11 5
norm_forward_asc.py
in tile_kernels/quant
 379 11 4
swiglu_forward_cuda.py
in tile_kernels/quant
 260 2 4
78 1 4
pre_split_mixes_cuda.py
in tile_kernels/mhc
 177 3 4
148 1 4
swiglu_backward_kernel.py
in tile_kernels/quant
 180 3 3
per_channel_cast_cuda.py
in tile_kernels/quant
 136 2 3
per_token_cast_kernel.py
in tile_kernels/quant
 149 4 3
102 1 3
engram_gate_fwd_cuda.py
in tile_kernels/engram
 163 4 3
benchmark.py
in tile_kernels/testing/pytest
 656 45 3
generator.py
in tile_kernels/testing
 131 11 3
rope_cuda.py
in tile_kernels/transform
 115 2 3
59 1 3
220 3 3
swiglu_forward_kernel.py
in tile_kernels/quant
 120 3 2
norm_backward_asc.py
in tile_kernels/quant
 172 5 2
common.py
in tile_kernels/quant
 261 24 2
72 1 2
per_block_cast_kernel.py
in tile_kernels/quant
 134 4 2
engram_hash_asc.py
in tile_kernels/engram
 77 2 2
engram.py
in tile_kernels/torch
 108 6 2
topk.py
in tile_kernels/torch
 141 4 2
pre_big_fuse_cuda.py
in tile_kernels/mhc
 106 1 2
sinkhorn_kernel.py
in tile_kernels/mhc
 15 2 2
Correlations

File Size vs. Commits (all time): 138 points

.pre-commit-config.yaml x: 2 commits (all time) y: 12 lines of code tile_kernels/__init__.py x: 2 commits (all time) y: 36 lines of code tile_kernels/config.py x: 2 commits (all time) y: 90 lines of code tile_kernels/engram/engram_fused_weight_asc.py x: 2 commits (all time) y: 49 lines of code tile_kernels/engram/engram_fused_weight_cuda.py x: 2 commits (all time) y: 37 lines of code tile_kernels/engram/engram_fused_weight_kernel.py x: 2 commits (all time) y: 21 lines of code tile_kernels/engram/engram_gate_bwd_asc.py x: 2 commits (all time) y: 181 lines of code tile_kernels/engram/engram_gate_bwd_cuda.py x: 2 commits (all time) y: 274 lines of code tile_kernels/engram/engram_gate_bwd_kernel.py x: 2 commits (all time) y: 61 lines of code tile_kernels/engram/engram_gate_fwd_asc.py x: 2 commits (all time) y: 233 lines of code tile_kernels/engram/engram_gate_fwd_cuda.py x: 2 commits (all time) y: 163 lines of code tile_kernels/engram/engram_gate_fwd_kernel.py x: 2 commits (all time) y: 62 lines of code tile_kernels/engram/engram_grad_w_reduce_asc.py x: 2 commits (all time) y: 66 lines of code tile_kernels/engram/engram_grad_w_reduce_cuda.py x: 2 commits (all time) y: 56 lines of code tile_kernels/engram/engram_grad_w_reduce_kernel.py x: 2 commits (all time) y: 31 lines of code tile_kernels/engram/engram_hash_asc.py x: 2 commits (all time) y: 77 lines of code tile_kernels/engram/engram_sinkhorn_asc.py x: 2 commits (all time) y: 252 lines of code tile_kernels/engram/engram_sinkhorn_cuda.py x: 2 commits (all time) y: 213 lines of code tile_kernels/engram/engram_sinkhorn_kernel.py x: 2 commits (all time) y: 112 lines of code tile_kernels/mhc/__init__.py x: 2 commits (all time) y: 17 lines of code tile_kernels/mhc/expand_cuda.py x: 2 commits (all time) y: 45 lines of code tile_kernels/mhc/head_compute_mix_asc.py x: 2 commits (all time) y: 85 lines of code tile_kernels/mhc/head_compute_mix_kernel.py x: 2 commits (all time) y: 94 lines of code tile_kernels/mhc/multilayer_recompute_asc.py x: 2 commits (all time) y: 138 lines of code tile_kernels/mhc/multilayer_recompute_cuda.py x: 2 commits (all time) y: 103 lines of code tile_kernels/mhc/norm_fn_asc.py x: 2 commits (all time) y: 369 lines of code tile_kernels/mhc/norm_fn_cuda.py x: 2 commits (all time) y: 131 lines of code tile_kernels/mhc/norm_fn_kernel.py x: 2 commits (all time) y: 117 lines of code tile_kernels/mhc/post_cuda.py x: 2 commits (all time) y: 127 lines of code tile_kernels/mhc/pre_apply_mix_asc.py x: 2 commits (all time) y: 99 lines of code tile_kernels/mhc/pre_apply_mix_cuda.py x: 2 commits (all time) y: 92 lines of code tile_kernels/mhc/pre_big_fuse_asc.py x: 2 commits (all time) y: 354 lines of code tile_kernels/mhc/pre_big_fuse_cuda.py x: 2 commits (all time) y: 106 lines of code tile_kernels/mhc/pre_split_mixes_asc.py x: 2 commits (all time) y: 414 lines of code tile_kernels/mhc/pre_split_mixes_cuda.py x: 2 commits (all time) y: 177 lines of code tile_kernels/mhc/sinkhorn_asc.py x: 2 commits (all time) y: 245 lines of code tile_kernels/mhc/sinkhorn_kernel.py x: 2 commits (all time) y: 15 lines of code tile_kernels/modeling/__init__.py x: 2 commits (all time) y: 1 lines of code tile_kernels/moe/__init__.py x: 2 commits (all time) y: 4 lines of code tile_kernels/moe/moe_topk_gate_backward_asc.py x: 2 commits (all time) y: 148 lines of code tile_kernels/moe/moe_topk_gate_backward_kernel.py x: 2 commits (all time) y: 59 lines of code tile_kernels/moe/moe_topk_gate_forward_asc.py x: 2 commits (all time) y: 265 lines of code tile_kernels/moe/moe_topk_gate_forward_cuda.py x: 2 commits (all time) y: 220 lines of code tile_kernels/moe/moe_topk_gate_forward_kernel.py x: 2 commits (all time) y: 110 lines of code tile_kernels/moe/normalize_weight_cuda.py x: 2 commits (all time) y: 41 lines of code tile_kernels/quant/batched_transpose_weight_sf_cuda.py x: 2 commits (all time) y: 78 lines of code tile_kernels/quant/batched_transpose_weight_sf_kernel.py x: 2 commits (all time) y: 72 lines of code tile_kernels/quant/cast_back_asc.py x: 2 commits (all time) y: 190 lines of code tile_kernels/quant/common.py x: 2 commits (all time) y: 261 lines of code tile_kernels/quant/norm_backward_asc.py x: 2 commits (all time) y: 172 lines of code tile_kernels/quant/norm_backward_cuda.py x: 2 commits (all time) y: 192 lines of code tile_kernels/quant/norm_forward_asc.py x: 2 commits (all time) y: 379 lines of code tile_kernels/quant/norm_forward_cuda.py x: 2 commits (all time) y: 217 lines of code tile_kernels/quant/norm_forward_kernel.py x: 2 commits (all time) y: 194 lines of code tile_kernels/quant/per_block_cast_cuda.py x: 2 commits (all time) y: 122 lines of code tile_kernels/quant/per_block_cast_kernel.py x: 2 commits (all time) y: 134 lines of code tile_kernels/quant/per_channel_cast_cuda.py x: 2 commits (all time) y: 136 lines of code tile_kernels/quant/per_token_cast_and_cast_back_asc.py x: 2 commits (all time) y: 330 lines of code tile_kernels/quant/per_token_cast_and_cast_back_cuda.py x: 2 commits (all time) y: 68 lines of code tile_kernels/quant/per_token_cast_asc.py x: 2 commits (all time) y: 336 lines of code tile_kernels/quant/per_token_cast_cuda.py x: 2 commits (all time) y: 179 lines of code tile_kernels/quant/swiglu_backward_asc.py x: 2 commits (all time) y: 445 lines of code tile_kernels/quant/swiglu_backward_cuda.py x: 2 commits (all time) y: 211 lines of code tile_kernels/quant/swiglu_forward_asc.py x: 2 commits (all time) y: 347 lines of code tile_kernels/quant/swiglu_forward_cuda.py x: 2 commits (all time) y: 260 lines of code tile_kernels/quant/swiglu_forward_kernel.py x: 2 commits (all time) y: 120 lines of code tile_kernels/testing/moe.py x: 2 commits (all time) y: 27 lines of code tile_kernels/testing/numeric.py x: 2 commits (all time) y: 52 lines of code tile_kernels/testing/pytest/benchmark.py x: 2 commits (all time) y: 656 lines of code tile_kernels/testing/pytest/common.py x: 2 commits (all time) y: 25 lines of code tile_kernels/testing/pytest/gpu_mem.py x: 2 commits (all time) y: 319 lines of code tile_kernels/torch/cast.py x: 2 commits (all time) y: 255 lines of code tile_kernels/torch/mhc.py x: 2 commits (all time) y: 86 lines of code tile_kernels/torch/rope.py x: 2 commits (all time) y: 42 lines of code tile_kernels/torch/swiglu.py x: 2 commits (all time) y: 143 lines of code tile_kernels/torch/topk.py x: 2 commits (all time) y: 141 lines of code tile_kernels/transform/rope_asc.py x: 2 commits (all time) y: 185 lines of code tile_kernels/transform/rope_cuda.py x: 2 commits (all time) y: 115 lines of code
656.0
lines of code
  min: 1.0
  average: 117.3
  25th percentile: 38.75
  median: 85.5
  75th percentile: 166.0
  max: 656.0
0 2.0
commits (all time)
min: 2.0 | average: 2.0 | 25th percentile: 2.0 | median: 2.0 | 75th percentile: 2.0 | max: 2.0

File Size vs. Contributors (all time): 138 points

.pre-commit-config.yaml x: 2 contributors (all time) y: 12 lines of code tile_kernels/__init__.py x: 2 contributors (all time) y: 36 lines of code tile_kernels/config.py x: 2 contributors (all time) y: 90 lines of code tile_kernels/engram/engram_fused_weight_asc.py x: 2 contributors (all time) y: 49 lines of code tile_kernels/engram/engram_fused_weight_cuda.py x: 2 contributors (all time) y: 37 lines of code tile_kernels/engram/engram_fused_weight_kernel.py x: 2 contributors (all time) y: 21 lines of code tile_kernels/engram/engram_gate_bwd_asc.py x: 2 contributors (all time) y: 181 lines of code tile_kernels/engram/engram_gate_bwd_cuda.py x: 2 contributors (all time) y: 274 lines of code tile_kernels/engram/engram_gate_bwd_kernel.py x: 2 contributors (all time) y: 61 lines of code tile_kernels/engram/engram_gate_fwd_asc.py x: 2 contributors (all time) y: 233 lines of code tile_kernels/engram/engram_gate_fwd_cuda.py x: 2 contributors (all time) y: 163 lines of code tile_kernels/engram/engram_gate_fwd_kernel.py x: 2 contributors (all time) y: 62 lines of code tile_kernels/engram/engram_grad_w_reduce_asc.py x: 2 contributors (all time) y: 66 lines of code tile_kernels/engram/engram_grad_w_reduce_cuda.py x: 2 contributors (all time) y: 56 lines of code tile_kernels/engram/engram_grad_w_reduce_kernel.py x: 2 contributors (all time) y: 31 lines of code tile_kernels/engram/engram_hash_asc.py x: 2 contributors (all time) y: 77 lines of code tile_kernels/engram/engram_sinkhorn_asc.py x: 2 contributors (all time) y: 252 lines of code tile_kernels/engram/engram_sinkhorn_cuda.py x: 2 contributors (all time) y: 213 lines of code tile_kernels/engram/engram_sinkhorn_kernel.py x: 2 contributors (all time) y: 112 lines of code tile_kernels/mhc/__init__.py x: 2 contributors (all time) y: 17 lines of code tile_kernels/mhc/expand_cuda.py x: 2 contributors (all time) y: 45 lines of code tile_kernels/mhc/head_compute_mix_asc.py x: 2 contributors (all time) y: 85 lines of code tile_kernels/mhc/head_compute_mix_kernel.py x: 2 contributors (all time) y: 94 lines of code tile_kernels/mhc/multilayer_recompute_asc.py x: 2 contributors (all time) y: 138 lines of code tile_kernels/mhc/multilayer_recompute_cuda.py x: 2 contributors (all time) y: 103 lines of code tile_kernels/mhc/norm_fn_asc.py x: 2 contributors (all time) y: 369 lines of code tile_kernels/mhc/norm_fn_cuda.py x: 2 contributors (all time) y: 131 lines of code tile_kernels/mhc/norm_fn_kernel.py x: 2 contributors (all time) y: 117 lines of code tile_kernels/mhc/post_cuda.py x: 2 contributors (all time) y: 127 lines of code tile_kernels/mhc/pre_apply_mix_asc.py x: 2 contributors (all time) y: 99 lines of code tile_kernels/mhc/pre_apply_mix_cuda.py x: 2 contributors (all time) y: 92 lines of code tile_kernels/mhc/pre_big_fuse_asc.py x: 2 contributors (all time) y: 354 lines of code tile_kernels/mhc/pre_big_fuse_cuda.py x: 2 contributors (all time) y: 106 lines of code tile_kernels/mhc/pre_split_mixes_asc.py x: 2 contributors (all time) y: 414 lines of code tile_kernels/mhc/pre_split_mixes_cuda.py x: 2 contributors (all time) y: 177 lines of code tile_kernels/mhc/sinkhorn_asc.py x: 2 contributors (all time) y: 245 lines of code tile_kernels/mhc/sinkhorn_kernel.py x: 2 contributors (all time) y: 15 lines of code tile_kernels/modeling/__init__.py x: 2 contributors (all time) y: 1 lines of code tile_kernels/moe/__init__.py x: 2 contributors (all time) y: 4 lines of code tile_kernels/moe/moe_topk_gate_backward_asc.py x: 2 contributors (all time) y: 148 lines of code tile_kernels/moe/moe_topk_gate_backward_kernel.py x: 2 contributors (all time) y: 59 lines of code tile_kernels/moe/moe_topk_gate_forward_asc.py x: 2 contributors (all time) y: 265 lines of code tile_kernels/moe/moe_topk_gate_forward_cuda.py x: 2 contributors (all time) y: 220 lines of code tile_kernels/moe/moe_topk_gate_forward_kernel.py x: 2 contributors (all time) y: 110 lines of code tile_kernels/moe/normalize_weight_cuda.py x: 2 contributors (all time) y: 41 lines of code tile_kernels/quant/batched_transpose_weight_sf_cuda.py x: 2 contributors (all time) y: 78 lines of code tile_kernels/quant/batched_transpose_weight_sf_kernel.py x: 2 contributors (all time) y: 72 lines of code tile_kernels/quant/cast_back_asc.py x: 2 contributors (all time) y: 190 lines of code tile_kernels/quant/common.py x: 2 contributors (all time) y: 261 lines of code tile_kernels/quant/norm_backward_asc.py x: 2 contributors (all time) y: 172 lines of code tile_kernels/quant/norm_backward_cuda.py x: 2 contributors (all time) y: 192 lines of code tile_kernels/quant/norm_forward_asc.py x: 2 contributors (all time) y: 379 lines of code tile_kernels/quant/norm_forward_cuda.py x: 2 contributors (all time) y: 217 lines of code tile_kernels/quant/norm_forward_kernel.py x: 2 contributors (all time) y: 194 lines of code tile_kernels/quant/per_block_cast_cuda.py x: 2 contributors (all time) y: 122 lines of code tile_kernels/quant/per_block_cast_kernel.py x: 2 contributors (all time) y: 134 lines of code tile_kernels/quant/per_channel_cast_cuda.py x: 2 contributors (all time) y: 136 lines of code tile_kernels/quant/per_token_cast_and_cast_back_asc.py x: 2 contributors (all time) y: 330 lines of code tile_kernels/quant/per_token_cast_and_cast_back_cuda.py x: 2 contributors (all time) y: 68 lines of code tile_kernels/quant/per_token_cast_asc.py x: 2 contributors (all time) y: 336 lines of code tile_kernels/quant/per_token_cast_cuda.py x: 2 contributors (all time) y: 179 lines of code tile_kernels/quant/swiglu_backward_asc.py x: 2 contributors (all time) y: 445 lines of code tile_kernels/quant/swiglu_backward_cuda.py x: 2 contributors (all time) y: 211 lines of code tile_kernels/quant/swiglu_forward_asc.py x: 2 contributors (all time) y: 347 lines of code tile_kernels/quant/swiglu_forward_cuda.py x: 2 contributors (all time) y: 260 lines of code tile_kernels/quant/swiglu_forward_kernel.py x: 2 contributors (all time) y: 120 lines of code tile_kernels/testing/moe.py x: 2 contributors (all time) y: 27 lines of code tile_kernels/testing/numeric.py x: 2 contributors (all time) y: 52 lines of code tile_kernels/testing/pytest/benchmark.py x: 2 contributors (all time) y: 656 lines of code tile_kernels/testing/pytest/common.py x: 2 contributors (all time) y: 25 lines of code tile_kernels/testing/pytest/gpu_mem.py x: 2 contributors (all time) y: 319 lines of code tile_kernels/torch/cast.py x: 2 contributors (all time) y: 255 lines of code tile_kernels/torch/mhc.py x: 2 contributors (all time) y: 86 lines of code tile_kernels/torch/rope.py x: 2 contributors (all time) y: 42 lines of code tile_kernels/torch/swiglu.py x: 2 contributors (all time) y: 143 lines of code tile_kernels/torch/topk.py x: 2 contributors (all time) y: 141 lines of code tile_kernels/transform/rope_asc.py x: 2 contributors (all time) y: 185 lines of code tile_kernels/transform/rope_cuda.py x: 2 contributors (all time) y: 115 lines of code
656.0
lines of code
  min: 1.0
  average: 117.3
  25th percentile: 38.75
  median: 85.5
  75th percentile: 166.0
  max: 656.0
0 2.0
contributors (all time)
min: 2.0 | average: 2.0 | 25th percentile: 2.0 | median: 2.0 | 75th percentile: 2.0 | max: 2.0

File Size vs. Commits (30 days): 138 points

.pre-commit-config.yaml x: 2 commits (30d) y: 12 lines of code tile_kernels/__init__.py x: 2 commits (30d) y: 36 lines of code tile_kernels/config.py x: 2 commits (30d) y: 90 lines of code tile_kernels/engram/engram_fused_weight_asc.py x: 2 commits (30d) y: 49 lines of code tile_kernels/engram/engram_fused_weight_cuda.py x: 2 commits (30d) y: 37 lines of code tile_kernels/engram/engram_fused_weight_kernel.py x: 2 commits (30d) y: 21 lines of code tile_kernels/engram/engram_gate_bwd_asc.py x: 2 commits (30d) y: 181 lines of code tile_kernels/engram/engram_gate_bwd_cuda.py x: 2 commits (30d) y: 274 lines of code tile_kernels/engram/engram_gate_bwd_kernel.py x: 2 commits (30d) y: 61 lines of code tile_kernels/engram/engram_gate_fwd_asc.py x: 2 commits (30d) y: 233 lines of code tile_kernels/engram/engram_gate_fwd_cuda.py x: 2 commits (30d) y: 163 lines of code tile_kernels/engram/engram_gate_fwd_kernel.py x: 2 commits (30d) y: 62 lines of code tile_kernels/engram/engram_grad_w_reduce_asc.py x: 2 commits (30d) y: 66 lines of code tile_kernels/engram/engram_grad_w_reduce_cuda.py x: 2 commits (30d) y: 56 lines of code tile_kernels/engram/engram_grad_w_reduce_kernel.py x: 2 commits (30d) y: 31 lines of code tile_kernels/engram/engram_hash_asc.py x: 2 commits (30d) y: 77 lines of code tile_kernels/engram/engram_sinkhorn_asc.py x: 2 commits (30d) y: 252 lines of code tile_kernels/engram/engram_sinkhorn_cuda.py x: 2 commits (30d) y: 213 lines of code tile_kernels/engram/engram_sinkhorn_kernel.py x: 2 commits (30d) y: 112 lines of code tile_kernels/mhc/__init__.py x: 2 commits (30d) y: 17 lines of code tile_kernels/mhc/expand_cuda.py x: 2 commits (30d) y: 45 lines of code tile_kernels/mhc/head_compute_mix_asc.py x: 2 commits (30d) y: 85 lines of code tile_kernels/mhc/head_compute_mix_kernel.py x: 2 commits (30d) y: 94 lines of code tile_kernels/mhc/multilayer_recompute_asc.py x: 2 commits (30d) y: 138 lines of code tile_kernels/mhc/multilayer_recompute_cuda.py x: 2 commits (30d) y: 103 lines of code tile_kernels/mhc/norm_fn_asc.py x: 2 commits (30d) y: 369 lines of code tile_kernels/mhc/norm_fn_cuda.py x: 2 commits (30d) y: 131 lines of code tile_kernels/mhc/norm_fn_kernel.py x: 2 commits (30d) y: 117 lines of code tile_kernels/mhc/post_cuda.py x: 2 commits (30d) y: 127 lines of code tile_kernels/mhc/pre_apply_mix_asc.py x: 2 commits (30d) y: 99 lines of code tile_kernels/mhc/pre_apply_mix_cuda.py x: 2 commits (30d) y: 92 lines of code tile_kernels/mhc/pre_big_fuse_asc.py x: 2 commits (30d) y: 354 lines of code tile_kernels/mhc/pre_big_fuse_cuda.py x: 2 commits (30d) y: 106 lines of code tile_kernels/mhc/pre_split_mixes_asc.py x: 2 commits (30d) y: 414 lines of code tile_kernels/mhc/pre_split_mixes_cuda.py x: 2 commits (30d) y: 177 lines of code tile_kernels/mhc/sinkhorn_asc.py x: 2 commits (30d) y: 245 lines of code tile_kernels/mhc/sinkhorn_kernel.py x: 2 commits (30d) y: 15 lines of code tile_kernels/modeling/__init__.py x: 2 commits (30d) y: 1 lines of code tile_kernels/moe/__init__.py x: 2 commits (30d) y: 4 lines of code tile_kernels/moe/moe_topk_gate_backward_asc.py x: 2 commits (30d) y: 148 lines of code tile_kernels/moe/moe_topk_gate_backward_kernel.py x: 2 commits (30d) y: 59 lines of code tile_kernels/moe/moe_topk_gate_forward_asc.py x: 2 commits (30d) y: 265 lines of code tile_kernels/moe/moe_topk_gate_forward_cuda.py x: 2 commits (30d) y: 220 lines of code tile_kernels/moe/moe_topk_gate_forward_kernel.py x: 2 commits (30d) y: 110 lines of code tile_kernels/moe/normalize_weight_cuda.py x: 2 commits (30d) y: 41 lines of code tile_kernels/quant/batched_transpose_weight_sf_cuda.py x: 2 commits (30d) y: 78 lines of code tile_kernels/quant/batched_transpose_weight_sf_kernel.py x: 2 commits (30d) y: 72 lines of code tile_kernels/quant/cast_back_asc.py x: 2 commits (30d) y: 190 lines of code tile_kernels/quant/common.py x: 2 commits (30d) y: 261 lines of code tile_kernels/quant/norm_backward_asc.py x: 2 commits (30d) y: 172 lines of code tile_kernels/quant/norm_backward_cuda.py x: 2 commits (30d) y: 192 lines of code tile_kernels/quant/norm_forward_asc.py x: 2 commits (30d) y: 379 lines of code tile_kernels/quant/norm_forward_cuda.py x: 2 commits (30d) y: 217 lines of code tile_kernels/quant/norm_forward_kernel.py x: 2 commits (30d) y: 194 lines of code tile_kernels/quant/per_block_cast_cuda.py x: 2 commits (30d) y: 122 lines of code tile_kernels/quant/per_block_cast_kernel.py x: 2 commits (30d) y: 134 lines of code tile_kernels/quant/per_channel_cast_cuda.py x: 2 commits (30d) y: 136 lines of code tile_kernels/quant/per_token_cast_and_cast_back_asc.py x: 2 commits (30d) y: 330 lines of code tile_kernels/quant/per_token_cast_and_cast_back_cuda.py x: 2 commits (30d) y: 68 lines of code tile_kernels/quant/per_token_cast_asc.py x: 2 commits (30d) y: 336 lines of code tile_kernels/quant/per_token_cast_cuda.py x: 2 commits (30d) y: 179 lines of code tile_kernels/quant/swiglu_backward_asc.py x: 2 commits (30d) y: 445 lines of code tile_kernels/quant/swiglu_backward_cuda.py x: 2 commits (30d) y: 211 lines of code tile_kernels/quant/swiglu_forward_asc.py x: 2 commits (30d) y: 347 lines of code tile_kernels/quant/swiglu_forward_cuda.py x: 2 commits (30d) y: 260 lines of code tile_kernels/quant/swiglu_forward_kernel.py x: 2 commits (30d) y: 120 lines of code tile_kernels/testing/moe.py x: 2 commits (30d) y: 27 lines of code tile_kernels/testing/numeric.py x: 2 commits (30d) y: 52 lines of code tile_kernels/testing/pytest/benchmark.py x: 2 commits (30d) y: 656 lines of code tile_kernels/testing/pytest/common.py x: 2 commits (30d) y: 25 lines of code tile_kernels/testing/pytest/gpu_mem.py x: 2 commits (30d) y: 319 lines of code tile_kernels/torch/cast.py x: 2 commits (30d) y: 255 lines of code tile_kernels/torch/mhc.py x: 2 commits (30d) y: 86 lines of code tile_kernels/torch/rope.py x: 2 commits (30d) y: 42 lines of code tile_kernels/torch/swiglu.py x: 2 commits (30d) y: 143 lines of code tile_kernels/torch/topk.py x: 2 commits (30d) y: 141 lines of code tile_kernels/transform/rope_asc.py x: 2 commits (30d) y: 185 lines of code tile_kernels/transform/rope_cuda.py x: 2 commits (30d) y: 115 lines of code
656.0
lines of code
  min: 1.0
  average: 117.3
  25th percentile: 38.75
  median: 85.5
  75th percentile: 166.0
  max: 656.0
0 2.0
commits (30d)
min: 2.0 | average: 2.0 | 25th percentile: 2.0 | median: 2.0 | 75th percentile: 2.0 | max: 2.0

File Size vs. Contributors (30 days): 138 points

.pre-commit-config.yaml x: 2 contributors (30d) y: 12 lines of code tile_kernels/__init__.py x: 2 contributors (30d) y: 36 lines of code tile_kernels/config.py x: 2 contributors (30d) y: 90 lines of code tile_kernels/engram/engram_fused_weight_asc.py x: 2 contributors (30d) y: 49 lines of code tile_kernels/engram/engram_fused_weight_cuda.py x: 2 contributors (30d) y: 37 lines of code tile_kernels/engram/engram_fused_weight_kernel.py x: 2 contributors (30d) y: 21 lines of code tile_kernels/engram/engram_gate_bwd_asc.py x: 2 contributors (30d) y: 181 lines of code tile_kernels/engram/engram_gate_bwd_cuda.py x: 2 contributors (30d) y: 274 lines of code tile_kernels/engram/engram_gate_bwd_kernel.py x: 2 contributors (30d) y: 61 lines of code tile_kernels/engram/engram_gate_fwd_asc.py x: 2 contributors (30d) y: 233 lines of code tile_kernels/engram/engram_gate_fwd_cuda.py x: 2 contributors (30d) y: 163 lines of code tile_kernels/engram/engram_gate_fwd_kernel.py x: 2 contributors (30d) y: 62 lines of code tile_kernels/engram/engram_grad_w_reduce_asc.py x: 2 contributors (30d) y: 66 lines of code tile_kernels/engram/engram_grad_w_reduce_cuda.py x: 2 contributors (30d) y: 56 lines of code tile_kernels/engram/engram_grad_w_reduce_kernel.py x: 2 contributors (30d) y: 31 lines of code tile_kernels/engram/engram_hash_asc.py x: 2 contributors (30d) y: 77 lines of code tile_kernels/engram/engram_sinkhorn_asc.py x: 2 contributors (30d) y: 252 lines of code tile_kernels/engram/engram_sinkhorn_cuda.py x: 2 contributors (30d) y: 213 lines of code tile_kernels/engram/engram_sinkhorn_kernel.py x: 2 contributors (30d) y: 112 lines of code tile_kernels/mhc/__init__.py x: 2 contributors (30d) y: 17 lines of code tile_kernels/mhc/expand_cuda.py x: 2 contributors (30d) y: 45 lines of code tile_kernels/mhc/head_compute_mix_asc.py x: 2 contributors (30d) y: 85 lines of code tile_kernels/mhc/head_compute_mix_kernel.py x: 2 contributors (30d) y: 94 lines of code tile_kernels/mhc/multilayer_recompute_asc.py x: 2 contributors (30d) y: 138 lines of code tile_kernels/mhc/multilayer_recompute_cuda.py x: 2 contributors (30d) y: 103 lines of code tile_kernels/mhc/norm_fn_asc.py x: 2 contributors (30d) y: 369 lines of code tile_kernels/mhc/norm_fn_cuda.py x: 2 contributors (30d) y: 131 lines of code tile_kernels/mhc/norm_fn_kernel.py x: 2 contributors (30d) y: 117 lines of code tile_kernels/mhc/post_cuda.py x: 2 contributors (30d) y: 127 lines of code tile_kernels/mhc/pre_apply_mix_asc.py x: 2 contributors (30d) y: 99 lines of code tile_kernels/mhc/pre_apply_mix_cuda.py x: 2 contributors (30d) y: 92 lines of code tile_kernels/mhc/pre_big_fuse_asc.py x: 2 contributors (30d) y: 354 lines of code tile_kernels/mhc/pre_big_fuse_cuda.py x: 2 contributors (30d) y: 106 lines of code tile_kernels/mhc/pre_split_mixes_asc.py x: 2 contributors (30d) y: 414 lines of code tile_kernels/mhc/pre_split_mixes_cuda.py x: 2 contributors (30d) y: 177 lines of code tile_kernels/mhc/sinkhorn_asc.py x: 2 contributors (30d) y: 245 lines of code tile_kernels/mhc/sinkhorn_kernel.py x: 2 contributors (30d) y: 15 lines of code tile_kernels/modeling/__init__.py x: 2 contributors (30d) y: 1 lines of code tile_kernels/moe/__init__.py x: 2 contributors (30d) y: 4 lines of code tile_kernels/moe/moe_topk_gate_backward_asc.py x: 2 contributors (30d) y: 148 lines of code tile_kernels/moe/moe_topk_gate_backward_kernel.py x: 2 contributors (30d) y: 59 lines of code tile_kernels/moe/moe_topk_gate_forward_asc.py x: 2 contributors (30d) y: 265 lines of code tile_kernels/moe/moe_topk_gate_forward_cuda.py x: 2 contributors (30d) y: 220 lines of code tile_kernels/moe/moe_topk_gate_forward_kernel.py x: 2 contributors (30d) y: 110 lines of code tile_kernels/moe/normalize_weight_cuda.py x: 2 contributors (30d) y: 41 lines of code tile_kernels/quant/batched_transpose_weight_sf_cuda.py x: 2 contributors (30d) y: 78 lines of code tile_kernels/quant/batched_transpose_weight_sf_kernel.py x: 2 contributors (30d) y: 72 lines of code tile_kernels/quant/cast_back_asc.py x: 2 contributors (30d) y: 190 lines of code tile_kernels/quant/common.py x: 2 contributors (30d) y: 261 lines of code tile_kernels/quant/norm_backward_asc.py x: 2 contributors (30d) y: 172 lines of code tile_kernels/quant/norm_backward_cuda.py x: 2 contributors (30d) y: 192 lines of code tile_kernels/quant/norm_forward_asc.py x: 2 contributors (30d) y: 379 lines of code tile_kernels/quant/norm_forward_cuda.py x: 2 contributors (30d) y: 217 lines of code tile_kernels/quant/norm_forward_kernel.py x: 2 contributors (30d) y: 194 lines of code tile_kernels/quant/per_block_cast_cuda.py x: 2 contributors (30d) y: 122 lines of code tile_kernels/quant/per_block_cast_kernel.py x: 2 contributors (30d) y: 134 lines of code tile_kernels/quant/per_channel_cast_cuda.py x: 2 contributors (30d) y: 136 lines of code tile_kernels/quant/per_token_cast_and_cast_back_asc.py x: 2 contributors (30d) y: 330 lines of code tile_kernels/quant/per_token_cast_and_cast_back_cuda.py x: 2 contributors (30d) y: 68 lines of code tile_kernels/quant/per_token_cast_asc.py x: 2 contributors (30d) y: 336 lines of code tile_kernels/quant/per_token_cast_cuda.py x: 2 contributors (30d) y: 179 lines of code tile_kernels/quant/swiglu_backward_asc.py x: 2 contributors (30d) y: 445 lines of code tile_kernels/quant/swiglu_backward_cuda.py x: 2 contributors (30d) y: 211 lines of code tile_kernels/quant/swiglu_forward_asc.py x: 2 contributors (30d) y: 347 lines of code tile_kernels/quant/swiglu_forward_cuda.py x: 2 contributors (30d) y: 260 lines of code tile_kernels/quant/swiglu_forward_kernel.py x: 2 contributors (30d) y: 120 lines of code tile_kernels/testing/moe.py x: 2 contributors (30d) y: 27 lines of code tile_kernels/testing/numeric.py x: 2 contributors (30d) y: 52 lines of code tile_kernels/testing/pytest/benchmark.py x: 2 contributors (30d) y: 656 lines of code tile_kernels/testing/pytest/common.py x: 2 contributors (30d) y: 25 lines of code tile_kernels/testing/pytest/gpu_mem.py x: 2 contributors (30d) y: 319 lines of code tile_kernels/torch/cast.py x: 2 contributors (30d) y: 255 lines of code tile_kernels/torch/mhc.py x: 2 contributors (30d) y: 86 lines of code tile_kernels/torch/rope.py x: 2 contributors (30d) y: 42 lines of code tile_kernels/torch/swiglu.py x: 2 contributors (30d) y: 143 lines of code tile_kernels/torch/topk.py x: 2 contributors (30d) y: 141 lines of code tile_kernels/transform/rope_asc.py x: 2 contributors (30d) y: 185 lines of code tile_kernels/transform/rope_cuda.py x: 2 contributors (30d) y: 115 lines of code
656.0
lines of code
  min: 1.0
  average: 117.3
  25th percentile: 38.75
  median: 85.5
  75th percentile: 166.0
  max: 656.0
0 2.0
contributors (30d)
min: 2.0 | average: 2.0 | 25th percentile: 2.0 | median: 2.0 | 75th percentile: 2.0 | max: 2.0

File Size vs. Commits (90 days): 138 points

.pre-commit-config.yaml x: 2 commits (90d) y: 12 lines of code tile_kernels/__init__.py x: 2 commits (90d) y: 36 lines of code tile_kernels/config.py x: 2 commits (90d) y: 90 lines of code tile_kernels/engram/engram_fused_weight_asc.py x: 2 commits (90d) y: 49 lines of code tile_kernels/engram/engram_fused_weight_cuda.py x: 2 commits (90d) y: 37 lines of code tile_kernels/engram/engram_fused_weight_kernel.py x: 2 commits (90d) y: 21 lines of code tile_kernels/engram/engram_gate_bwd_asc.py x: 2 commits (90d) y: 181 lines of code tile_kernels/engram/engram_gate_bwd_cuda.py x: 2 commits (90d) y: 274 lines of code tile_kernels/engram/engram_gate_bwd_kernel.py x: 2 commits (90d) y: 61 lines of code tile_kernels/engram/engram_gate_fwd_asc.py x: 2 commits (90d) y: 233 lines of code tile_kernels/engram/engram_gate_fwd_cuda.py x: 2 commits (90d) y: 163 lines of code tile_kernels/engram/engram_gate_fwd_kernel.py x: 2 commits (90d) y: 62 lines of code tile_kernels/engram/engram_grad_w_reduce_asc.py x: 2 commits (90d) y: 66 lines of code tile_kernels/engram/engram_grad_w_reduce_cuda.py x: 2 commits (90d) y: 56 lines of code tile_kernels/engram/engram_grad_w_reduce_kernel.py x: 2 commits (90d) y: 31 lines of code tile_kernels/engram/engram_hash_asc.py x: 2 commits (90d) y: 77 lines of code tile_kernels/engram/engram_sinkhorn_asc.py x: 2 commits (90d) y: 252 lines of code tile_kernels/engram/engram_sinkhorn_cuda.py x: 2 commits (90d) y: 213 lines of code tile_kernels/engram/engram_sinkhorn_kernel.py x: 2 commits (90d) y: 112 lines of code tile_kernels/mhc/__init__.py x: 2 commits (90d) y: 17 lines of code tile_kernels/mhc/expand_cuda.py x: 2 commits (90d) y: 45 lines of code tile_kernels/mhc/head_compute_mix_asc.py x: 2 commits (90d) y: 85 lines of code tile_kernels/mhc/head_compute_mix_kernel.py x: 2 commits (90d) y: 94 lines of code tile_kernels/mhc/multilayer_recompute_asc.py x: 2 commits (90d) y: 138 lines of code tile_kernels/mhc/multilayer_recompute_cuda.py x: 2 commits (90d) y: 103 lines of code tile_kernels/mhc/norm_fn_asc.py x: 2 commits (90d) y: 369 lines of code tile_kernels/mhc/norm_fn_cuda.py x: 2 commits (90d) y: 131 lines of code tile_kernels/mhc/norm_fn_kernel.py x: 2 commits (90d) y: 117 lines of code tile_kernels/mhc/post_cuda.py x: 2 commits (90d) y: 127 lines of code tile_kernels/mhc/pre_apply_mix_asc.py x: 2 commits (90d) y: 99 lines of code tile_kernels/mhc/pre_apply_mix_cuda.py x: 2 commits (90d) y: 92 lines of code tile_kernels/mhc/pre_big_fuse_asc.py x: 2 commits (90d) y: 354 lines of code tile_kernels/mhc/pre_big_fuse_cuda.py x: 2 commits (90d) y: 106 lines of code tile_kernels/mhc/pre_split_mixes_asc.py x: 2 commits (90d) y: 414 lines of code tile_kernels/mhc/pre_split_mixes_cuda.py x: 2 commits (90d) y: 177 lines of code tile_kernels/mhc/sinkhorn_asc.py x: 2 commits (90d) y: 245 lines of code tile_kernels/mhc/sinkhorn_kernel.py x: 2 commits (90d) y: 15 lines of code tile_kernels/modeling/__init__.py x: 2 commits (90d) y: 1 lines of code tile_kernels/moe/__init__.py x: 2 commits (90d) y: 4 lines of code tile_kernels/moe/moe_topk_gate_backward_asc.py x: 2 commits (90d) y: 148 lines of code tile_kernels/moe/moe_topk_gate_backward_kernel.py x: 2 commits (90d) y: 59 lines of code tile_kernels/moe/moe_topk_gate_forward_asc.py x: 2 commits (90d) y: 265 lines of code tile_kernels/moe/moe_topk_gate_forward_cuda.py x: 2 commits (90d) y: 220 lines of code tile_kernels/moe/moe_topk_gate_forward_kernel.py x: 2 commits (90d) y: 110 lines of code tile_kernels/moe/normalize_weight_cuda.py x: 2 commits (90d) y: 41 lines of code tile_kernels/quant/batched_transpose_weight_sf_cuda.py x: 2 commits (90d) y: 78 lines of code tile_kernels/quant/batched_transpose_weight_sf_kernel.py x: 2 commits (90d) y: 72 lines of code tile_kernels/quant/cast_back_asc.py x: 2 commits (90d) y: 190 lines of code tile_kernels/quant/common.py x: 2 commits (90d) y: 261 lines of code tile_kernels/quant/norm_backward_asc.py x: 2 commits (90d) y: 172 lines of code tile_kernels/quant/norm_backward_cuda.py x: 2 commits (90d) y: 192 lines of code tile_kernels/quant/norm_forward_asc.py x: 2 commits (90d) y: 379 lines of code tile_kernels/quant/norm_forward_cuda.py x: 2 commits (90d) y: 217 lines of code tile_kernels/quant/norm_forward_kernel.py x: 2 commits (90d) y: 194 lines of code tile_kernels/quant/per_block_cast_cuda.py x: 2 commits (90d) y: 122 lines of code tile_kernels/quant/per_block_cast_kernel.py x: 2 commits (90d) y: 134 lines of code tile_kernels/quant/per_channel_cast_cuda.py x: 2 commits (90d) y: 136 lines of code tile_kernels/quant/per_token_cast_and_cast_back_asc.py x: 2 commits (90d) y: 330 lines of code tile_kernels/quant/per_token_cast_and_cast_back_cuda.py x: 2 commits (90d) y: 68 lines of code tile_kernels/quant/per_token_cast_asc.py x: 2 commits (90d) y: 336 lines of code tile_kernels/quant/per_token_cast_cuda.py x: 2 commits (90d) y: 179 lines of code tile_kernels/quant/swiglu_backward_asc.py x: 2 commits (90d) y: 445 lines of code tile_kernels/quant/swiglu_backward_cuda.py x: 2 commits (90d) y: 211 lines of code tile_kernels/quant/swiglu_forward_asc.py x: 2 commits (90d) y: 347 lines of code tile_kernels/quant/swiglu_forward_cuda.py x: 2 commits (90d) y: 260 lines of code tile_kernels/quant/swiglu_forward_kernel.py x: 2 commits (90d) y: 120 lines of code tile_kernels/testing/moe.py x: 2 commits (90d) y: 27 lines of code tile_kernels/testing/numeric.py x: 2 commits (90d) y: 52 lines of code tile_kernels/testing/pytest/benchmark.py x: 2 commits (90d) y: 656 lines of code tile_kernels/testing/pytest/common.py x: 2 commits (90d) y: 25 lines of code tile_kernels/testing/pytest/gpu_mem.py x: 2 commits (90d) y: 319 lines of code tile_kernels/torch/cast.py x: 2 commits (90d) y: 255 lines of code tile_kernels/torch/mhc.py x: 2 commits (90d) y: 86 lines of code tile_kernels/torch/rope.py x: 2 commits (90d) y: 42 lines of code tile_kernels/torch/swiglu.py x: 2 commits (90d) y: 143 lines of code tile_kernels/torch/topk.py x: 2 commits (90d) y: 141 lines of code tile_kernels/transform/rope_asc.py x: 2 commits (90d) y: 185 lines of code tile_kernels/transform/rope_cuda.py x: 2 commits (90d) y: 115 lines of code
656.0
lines of code
  min: 1.0
  average: 117.3
  25th percentile: 38.75
  median: 85.5
  75th percentile: 166.0
  max: 656.0
0 2.0
commits (90d)
min: 2.0 | average: 2.0 | 25th percentile: 2.0 | median: 2.0 | 75th percentile: 2.0 | max: 2.0

File Size vs. Contributors (90 days): 138 points

.pre-commit-config.yaml x: 2 contributors (90d) y: 12 lines of code tile_kernels/__init__.py x: 2 contributors (90d) y: 36 lines of code tile_kernels/config.py x: 2 contributors (90d) y: 90 lines of code tile_kernels/engram/engram_fused_weight_asc.py x: 2 contributors (90d) y: 49 lines of code tile_kernels/engram/engram_fused_weight_cuda.py x: 2 contributors (90d) y: 37 lines of code tile_kernels/engram/engram_fused_weight_kernel.py x: 2 contributors (90d) y: 21 lines of code tile_kernels/engram/engram_gate_bwd_asc.py x: 2 contributors (90d) y: 181 lines of code tile_kernels/engram/engram_gate_bwd_cuda.py x: 2 contributors (90d) y: 274 lines of code tile_kernels/engram/engram_gate_bwd_kernel.py x: 2 contributors (90d) y: 61 lines of code tile_kernels/engram/engram_gate_fwd_asc.py x: 2 contributors (90d) y: 233 lines of code tile_kernels/engram/engram_gate_fwd_cuda.py x: 2 contributors (90d) y: 163 lines of code tile_kernels/engram/engram_gate_fwd_kernel.py x: 2 contributors (90d) y: 62 lines of code tile_kernels/engram/engram_grad_w_reduce_asc.py x: 2 contributors (90d) y: 66 lines of code tile_kernels/engram/engram_grad_w_reduce_cuda.py x: 2 contributors (90d) y: 56 lines of code tile_kernels/engram/engram_grad_w_reduce_kernel.py x: 2 contributors (90d) y: 31 lines of code tile_kernels/engram/engram_hash_asc.py x: 2 contributors (90d) y: 77 lines of code tile_kernels/engram/engram_sinkhorn_asc.py x: 2 contributors (90d) y: 252 lines of code tile_kernels/engram/engram_sinkhorn_cuda.py x: 2 contributors (90d) y: 213 lines of code tile_kernels/engram/engram_sinkhorn_kernel.py x: 2 contributors (90d) y: 112 lines of code tile_kernels/mhc/__init__.py x: 2 contributors (90d) y: 17 lines of code tile_kernels/mhc/expand_cuda.py x: 2 contributors (90d) y: 45 lines of code tile_kernels/mhc/head_compute_mix_asc.py x: 2 contributors (90d) y: 85 lines of code tile_kernels/mhc/head_compute_mix_kernel.py x: 2 contributors (90d) y: 94 lines of code tile_kernels/mhc/multilayer_recompute_asc.py x: 2 contributors (90d) y: 138 lines of code tile_kernels/mhc/multilayer_recompute_cuda.py x: 2 contributors (90d) y: 103 lines of code tile_kernels/mhc/norm_fn_asc.py x: 2 contributors (90d) y: 369 lines of code tile_kernels/mhc/norm_fn_cuda.py x: 2 contributors (90d) y: 131 lines of code tile_kernels/mhc/norm_fn_kernel.py x: 2 contributors (90d) y: 117 lines of code tile_kernels/mhc/post_cuda.py x: 2 contributors (90d) y: 127 lines of code tile_kernels/mhc/pre_apply_mix_asc.py x: 2 contributors (90d) y: 99 lines of code tile_kernels/mhc/pre_apply_mix_cuda.py x: 2 contributors (90d) y: 92 lines of code tile_kernels/mhc/pre_big_fuse_asc.py x: 2 contributors (90d) y: 354 lines of code tile_kernels/mhc/pre_big_fuse_cuda.py x: 2 contributors (90d) y: 106 lines of code tile_kernels/mhc/pre_split_mixes_asc.py x: 2 contributors (90d) y: 414 lines of code tile_kernels/mhc/pre_split_mixes_cuda.py x: 2 contributors (90d) y: 177 lines of code tile_kernels/mhc/sinkhorn_asc.py x: 2 contributors (90d) y: 245 lines of code tile_kernels/mhc/sinkhorn_kernel.py x: 2 contributors (90d) y: 15 lines of code tile_kernels/modeling/__init__.py x: 2 contributors (90d) y: 1 lines of code tile_kernels/moe/__init__.py x: 2 contributors (90d) y: 4 lines of code tile_kernels/moe/moe_topk_gate_backward_asc.py x: 2 contributors (90d) y: 148 lines of code tile_kernels/moe/moe_topk_gate_backward_kernel.py x: 2 contributors (90d) y: 59 lines of code tile_kernels/moe/moe_topk_gate_forward_asc.py x: 2 contributors (90d) y: 265 lines of code tile_kernels/moe/moe_topk_gate_forward_cuda.py x: 2 contributors (90d) y: 220 lines of code tile_kernels/moe/moe_topk_gate_forward_kernel.py x: 2 contributors (90d) y: 110 lines of code tile_kernels/moe/normalize_weight_cuda.py x: 2 contributors (90d) y: 41 lines of code tile_kernels/quant/batched_transpose_weight_sf_cuda.py x: 2 contributors (90d) y: 78 lines of code tile_kernels/quant/batched_transpose_weight_sf_kernel.py x: 2 contributors (90d) y: 72 lines of code tile_kernels/quant/cast_back_asc.py x: 2 contributors (90d) y: 190 lines of code tile_kernels/quant/common.py x: 2 contributors (90d) y: 261 lines of code tile_kernels/quant/norm_backward_asc.py x: 2 contributors (90d) y: 172 lines of code tile_kernels/quant/norm_backward_cuda.py x: 2 contributors (90d) y: 192 lines of code tile_kernels/quant/norm_forward_asc.py x: 2 contributors (90d) y: 379 lines of code tile_kernels/quant/norm_forward_cuda.py x: 2 contributors (90d) y: 217 lines of code tile_kernels/quant/norm_forward_kernel.py x: 2 contributors (90d) y: 194 lines of code tile_kernels/quant/per_block_cast_cuda.py x: 2 contributors (90d) y: 122 lines of code tile_kernels/quant/per_block_cast_kernel.py x: 2 contributors (90d) y: 134 lines of code tile_kernels/quant/per_channel_cast_cuda.py x: 2 contributors (90d) y: 136 lines of code tile_kernels/quant/per_token_cast_and_cast_back_asc.py x: 2 contributors (90d) y: 330 lines of code tile_kernels/quant/per_token_cast_and_cast_back_cuda.py x: 2 contributors (90d) y: 68 lines of code tile_kernels/quant/per_token_cast_asc.py x: 2 contributors (90d) y: 336 lines of code tile_kernels/quant/per_token_cast_cuda.py x: 2 contributors (90d) y: 179 lines of code tile_kernels/quant/swiglu_backward_asc.py x: 2 contributors (90d) y: 445 lines of code tile_kernels/quant/swiglu_backward_cuda.py x: 2 contributors (90d) y: 211 lines of code tile_kernels/quant/swiglu_forward_asc.py x: 2 contributors (90d) y: 347 lines of code tile_kernels/quant/swiglu_forward_cuda.py x: 2 contributors (90d) y: 260 lines of code tile_kernels/quant/swiglu_forward_kernel.py x: 2 contributors (90d) y: 120 lines of code tile_kernels/testing/moe.py x: 2 contributors (90d) y: 27 lines of code tile_kernels/testing/numeric.py x: 2 contributors (90d) y: 52 lines of code tile_kernels/testing/pytest/benchmark.py x: 2 contributors (90d) y: 656 lines of code tile_kernels/testing/pytest/common.py x: 2 contributors (90d) y: 25 lines of code tile_kernels/testing/pytest/gpu_mem.py x: 2 contributors (90d) y: 319 lines of code tile_kernels/torch/cast.py x: 2 contributors (90d) y: 255 lines of code tile_kernels/torch/mhc.py x: 2 contributors (90d) y: 86 lines of code tile_kernels/torch/rope.py x: 2 contributors (90d) y: 42 lines of code tile_kernels/torch/swiglu.py x: 2 contributors (90d) y: 143 lines of code tile_kernels/torch/topk.py x: 2 contributors (90d) y: 141 lines of code tile_kernels/transform/rope_asc.py x: 2 contributors (90d) y: 185 lines of code tile_kernels/transform/rope_cuda.py x: 2 contributors (90d) y: 115 lines of code
656.0
lines of code
  min: 1.0
  average: 117.3
  25th percentile: 38.75
  median: 85.5
  75th percentile: 166.0
  max: 656.0
0 2.0
contributors (90d)
min: 2.0 | average: 2.0 | 25th percentile: 2.0 | median: 2.0 | 75th percentile: 2.0 | max: 2.0
About This Analysis
  • File size measurements show the distribution of size of files.
  • Files are classified in four categories based on their size (lines of code): 1-100 (very small files), 101-200 (small files), 201-500 (medium size files), 501-1000 (long files), 1001+(very long files).
  • It is a good practice to keep files small. Long files may become "bloaters", code that have increased to such gargantuan proportions that they are hard to work with.
Learn more...