triton_attn_tiny / debug.att.log
Yi30's picture
Add files using upload-large-folder tool
fbc96c0 verified
============================= test session starts ==============================
platform linux -- Python 3.12.3, pytest-8.1.1, pluggy-1.6.0 -- /usr/bin/python
cachedir: .pytest_cache
hypothesis profile 'default' -> database=DirectoryBasedExampleDatabase(PosixPath('/home/yiliu7/workspace/vllm/tests/kernels/attention/.hypothesis/examples'))
rootdir: /home/yiliu7/workspace/vllm
configfile: pyproject.toml
plugins: anyio-4.9.0, rerunfailures-15.1, shard-0.1.2, xdoctest-1.0.2, xdist-3.6.1, flakefinder-1.1.0, hypothesis-6.130.8, typeguard-4.3.0
collecting ... WARNING 11-29 09:56:20 [interface.py:508] Current platform cuda does not have '_pytestfixturefunction' attribute.
WARNING 11-29 09:56:20 [interface.py:508] Current platform cuda does not have '__test__' attribute.
WARNING 11-29 09:56:20 [interface.py:508] Current platform cuda does not have '__bases__' attribute.
WARNING 11-29 09:56:20 [interface.py:508] Current platform cuda does not have '__test__' attribute.
collected 1 item
Running 1 items in this shard: tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0] query shape: torch.Size([5, 8, 8])
query_idx:
tensor([[[ 0, 1, 2, 3, 4, 5, 6, 7],
[ 8, 9, 10, 11, 12, 13, 14, 15],
[ 16, 17, 18, 19, 20, 21, 22, 23],
[ 24, 25, 26, 27, 28, 29, 30, 31],
[ 32, 33, 34, 35, 36, 37, 38, 39],
[ 40, 41, 42, 43, 44, 45, 46, 47],
[ 48, 49, 50, 51, 52, 53, 54, 55],
[ 56, 57, 58, 59, 60, 61, 62, 63]],
[[ 64, 65, 66, 67, 68, 69, 70, 71],
[ 72, 73, 74, 75, 76, 77, 78, 79],
[ 80, 81, 82, 83, 84, 85, 86, 87],
[ 88, 89, 90, 91, 92, 93, 94, 95],
[ 96, 97, 98, 99, 100, 101, 102, 103],
[104, 105, 106, 107, 108, 109, 110, 111],
[112, 113, 114, 115, 116, 117, 118, 119],
[120, 121, 122, 123, 124, 125, 126, 127]],
[[128, 129, 130, 131, 132, 133, 134, 135],
[136, 137, 138, 139, 140, 141, 142, 143],
[144, 145, 146, 147, 148, 149, 150, 151],
[152, 153, 154, 155, 156, 157, 158, 159],
[160, 161, 162, 163, 164, 165, 166, 167],
[168, 169, 170, 171, 172, 173, 174, 175],
[176, 177, 178, 179, 180, 181, 182, 183],
[184, 185, 186, 187, 188, 189, 190, 191]],
[[192, 193, 194, 195, 196, 197, 198, 199],
[200, 201, 202, 203, 204, 205, 206, 207],
[208, 209, 210, 211, 212, 213, 214, 215],
[216, 217, 218, 219, 220, 221, 222, 223],
[224, 225, 226, 227, 228, 229, 230, 231],
[232, 233, 234, 235, 236, 237, 238, 239],
[240, 241, 242, 243, 244, 245, 246, 247],
[248, 249, 250, 251, 252, 253, 254, 255]],
[[256, 257, 258, 259, 260, 261, 262, 263],
[264, 265, 266, 267, 268, 269, 270, 271],
[272, 273, 274, 275, 276, 277, 278, 279],
[280, 281, 282, 283, 284, 285, 286, 287],
[288, 289, 290, 291, 292, 293, 294, 295],
[296, 297, 298, 299, 300, 301, 302, 303],
[304, 305, 306, 307, 308, 309, 310, 311],
[312, 313, 314, 315, 316, 317, 318, 319]]], device='cuda:0')
query_uint8:
tensor([[[183, 174, 195, 33, 160, 177, 178, 170],
[185, 172, 189, 65, 42, 164, 58, 57],
[160, 58, 60, 56, 176, 187, 180, 186],
[ 25, 132, 63, 51, 154, 181, 59, 169],
[178, 170, 183, 181, 180, 149, 49, 60],
[ 7, 46, 61, 51, 164, 56, 188, 180],
[164, 44, 45, 155, 54, 185, 30, 48],
[177, 170, 177, 54, 41, 32, 47, 185]],
[[145, 184, 55, 185, 184, 31, 51, 173],
[179, 177, 51, 188, 44, 147, 44, 176],
[170, 52, 196, 43, 142, 60, 129, 36],
[ 38, 167, 56, 61, 33, 48, 57, 183],
[155, 188, 182, 174, 24, 146, 49, 54],
[189, 46, 52, 182, 51, 171, 191, 163],
[ 63, 48, 57, 187, 168, 188, 60, 6],
[ 65, 183, 62, 168, 182, 166, 171, 165]],
[[185, 177, 178, 64, 187, 58, 192, 182],
[ 41, 180, 164, 171, 43, 161, 43, 184],
[173, 49, 21, 61, 173, 166, 163, 34],
[ 24, 187, 145, 167, 56, 58, 52, 181],
[ 45, 175, 184, 55, 44, 175, 189, 56],
[177, 41, 169, 34, 193, 49, 186, 60],
[ 29, 175, 47, 34, 187, 63, 185, 41],
[179, 51, 181, 156, 36, 38, 180, 174]],
[[161, 187, 182, 173, 44, 39, 187, 176],
[173, 188, 185, 168, 179, 193, 174, 23],
[ 41, 58, 136, 45, 186, 61, 48, 187],
[190, 55, 166, 177, 42, 175, 65, 49],
[ 55, 41, 193, 46, 57, 50, 42, 58],
[ 29, 192, 49, 34, 53, 40, 179, 134],
[194, 43, 43, 154, 38, 194, 25, 192],
[ 57, 177, 59, 190, 185, 59, 69, 182]],
[[ 39, 57, 168, 176, 185, 59, 55, 151],
[ 45, 175, 28, 57, 51, 164, 175, 190],
[ 25, 184, 162, 50, 45, 175, 55, 181],
[145, 186, 178, 57, 179, 57, 179, 161],
[ 51, 179, 36, 190, 177, 179, 160, 178],
[165, 62, 49, 152, 64, 168, 190, 174],
[185, 58, 43, 179, 174, 184, 183, 49],
[ 58, 191, 55, 180, 19, 63, 181, 179]]], device='cuda:0',
dtype=torch.uint8)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB set_trace >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/yiliu7/workspace/vllm/tests/kernels/attention/test_triton_unified_attention.py(219)test_triton_unified_attn()
-> ref_output = ref_paged_attn(
(Pdb)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB continue >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB set_trace >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/yiliu7/workspace/vllm/tests/kernels/attention/test_triton_unified_attention.py(56)ref_paged_attn()
-> for i in range(num_seqs):
(Pdb)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB continue >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
k shape: torch.Size([7, 2, 8])
k data:
tensor([[[186, 58, 189, 59, 179, 172, 49, 53],
[ 57, 54, 58, 51, 49, 134, 181, 63]],
[[168, 56, 35, 177, 66, 40, 43, 50],
[ 42, 157, 176, 184, 40, 57, 176, 181]],
[[186, 191, 54, 48, 180, 43, 38, 157],
[ 50, 164, 52, 189, 51, 64, 46, 191]],
[[ 51, 178, 180, 61, 179, 45, 60, 41],
[ 60, 52, 152, 182, 41, 173, 191, 178]],
[[132, 174, 50, 181, 180, 59, 180, 46],
[ 67, 42, 41, 43, 192, 64, 152, 186]],
[[170, 171, 184, 48, 43, 162, 169, 190],
[ 52, 179, 57, 50, 169, 168, 59, 40]],
[[ 47, 49, 57, 49, 57, 177, 178, 169],
[183, 189, 186, 164, 51, 33, 184, 48]]], device='cuda:0',
dtype=torch.uint8)
v shape: torch.Size([7, 2, 8])
v data:
tensor([[[181, 50, 178, 57, 177, 187, 59, 21],
[ 68, 57, 42, 179, 61, 50, 169, 147]],
[[173, 57, 39, 41, 42, 194, 183, 54],
[ 50, 188, 160, 42, 170, 173, 172, 170]],
[[ 17, 179, 181, 53, 151, 162, 60, 186],
[187, 182, 182, 41, 51, 37, 186, 183]],
[[177, 170, 39, 179, 188, 43, 59, 55],
[184, 64, 183, 53, 174, 160, 58, 194]],
[[ 34, 36, 191, 44, 32, 34, 50, 39],
[181, 59, 47, 184, 188, 50, 57, 61]],
[[ 46, 52, 147, 59, 43, 180, 21, 54],
[186, 36, 171, 49, 63, 61, 61, 129]],
[[ 1, 193, 157, 154, 183, 175, 43, 61],
[ 55, 54, 173, 56, 63, 60, 45, 56]]], device='cuda:0',
dtype=torch.uint8)
q shape : torch.Size([5, 8, 8]), k/v shape torch.Size([7, 8, 8])
attn shape: torch.Size([8, 5, 7]), v shape: torch.Size([7, 8, 8]), out shape: torch.Size([5, 8, 8])
num_seqs=1, num_query_heads=8, num_kv_heads=2
num_queries_per_kv=4, head_size=8
Using BLOCK_M=16, BLOCK_Q=4
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB set_trace >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/yiliu7/workspace/vllm/vllm/attention/ops/triton_unified_attention.py(762)unified_attention()
-> print(f"Launch parameters: total_num_q_blocks={total_num_q_blocks}, num_kv_heads={num_kv_heads}")
(Pdb)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB continue >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Launch parameters: total_num_q_blocks=2, num_kv_heads=2
q_block_global_idx: [0], kv_head_idx: [0]
q_block_global_idx: [0], kv_head_idx: [0]
query_offset_0:
[0 0 0 0 1 1 1 1 2 2 2 2 3 3 3 3]
query_offset_1:
[0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3]
query_offset:
[[ 0 1 2 3 4 5 6 7]
[ 8 9 10 11 12 13 14 15]
[ 16 17 18 19 20 21 22 23]
[ 24 25 26 27 28 29 30 31]
[ 64 65 66 67 68 69 70 71]
[ 72 73 74 75 76 77 78 79]
[ 80 81 82 83 84 85 86 87]
[ 88 89 90 91 92 93 94 95]
[128 129 130 131 132 133 134 135]
[136 137 138 139 140 141 142 143]
[144 145 146 147 148 149 150 151]
[152 153 154 155 156 157 158 159]
[192 193 194 195 196 197 198 199]
[200 201 202 203 204 205 206 207]
[208 209 210 211 212 213 214 215]
[216 217 218 219 220 221 222 223]]
Q:
[[183 174 195 33 160 177 178 170]
[185 172 189 65 42 164 58 57]
[160 58 60 56 176 187 180 186]
[ 25 132 63 51 154 181 59 169]
[145 184 55 185 184 31 51 173]
[179 177 51 188 44 147 44 176]
[170 52 196 43 142 60 129 36]
[ 38 167 56 61 33 48 57 183]
[185 177 178 64 187 58 192 182]
[ 41 180 164 171 43 161 43 184]
[173 49 21 61 173 166 163 34]
[ 24 187 145 167 56 58 52 181]
[161 187 182 173 44 39 187 176]
[173 188 185 168 179 193 174 23]
[ 41 58 136 45 186 61 48 187]
[190 55 166 177 42 175 65 49]]
q_block_global_idx: [0], kv_head_idx: [0]
k_offset:
[[469248 469264 469280 469296 469312 469328 469344 469360 469376 469392
469408 469424 469440 469456 469472 469488]
[469249 469265 469281 469297 469313 469329 469345 469361 469377 469393
469409 469425 469441 469457 469473 469489]
[469250 469266 469282 469298 469314 469330 469346 469362 469378 469394
469410 469426 469442 469458 469474 469490]
[469251 469267 469283 469299 469315 469331 469347 469363 469379 469395
469411 469427 469443 469459 469475 469491]
[469252 469268 469284 469300 469316 469332 469348 469364 469380 469396
469412 469428 469444 469460 469476 469492]
[469253 469269 469285 469301 469317 469333 469349 469365 469381 469397
469413 469429 469445 469461 469477 469493]
[469254 469270 469286 469302 469318 469334 469350 469366 469382 469398
469414 469430 469446 469462 469478 469494]
[469255 469271 469287 469303 469319 469335 469351 469367 469383 469399
469415 469431 469447 469463 469479 469495]]
K_load:
[[186 168 186 51 132 170 0 0 0 0 0 0 0 0 0 0]
[ 58 56 191 178 174 171 0 0 0 0 0 0 0 0 0 0]
[189 35 54 180 50 184 0 0 0 0 0 0 0 0 0 0]
[ 59 177 48 61 181 48 0 0 0 0 0 0 0 0 0 0]
[179 66 180 179 180 43 0 0 0 0 0 0 0 0 0 0]
[172 40 43 45 59 162 0 0 0 0 0 0 0 0 0 0]
[ 49 43 38 60 180 169 0 0 0 0 0 0 0 0 0 0]
[ 53 50 157 41 46 190 0 0 0 0 0 0 0 0 0 0]]
v_offset:
[[469248 469249 469250 469251 469252 469253 469254 469255]
[469264 469265 469266 469267 469268 469269 469270 469271]
[469280 469281 469282 469283 469284 469285 469286 469287]
[469296 469297 469298 469299 469300 469301 469302 469303]
[469312 469313 469314 469315 469316 469317 469318 469319]
[469328 469329 469330 469331 469332 469333 469334 469335]
[469344 469345 469346 469347 469348 469349 469350 469351]
[469360 469361 469362 469363 469364 469365 469366 469367]
[469376 469377 469378 469379 469380 469381 469382 469383]
[469392 469393 469394 469395 469396 469397 469398 469399]
[469408 469409 469410 469411 469412 469413 469414 469415]
[469424 469425 469426 469427 469428 469429 469430 469431]
[469440 469441 469442 469443 469444 469445 469446 469447]
[469456 469457 469458 469459 469460 469461 469462 469463]
[469472 469473 469474 469475 469476 469477 469478 469479]
[469488 469489 469490 469491 469492 469493 469494 469495]]
V_load:
[[181 50 178 57 177 187 59 21]
[173 57 39 41 42 194 183 54]
[ 17 179 181 53 151 162 60 186]
[177 170 39 179 188 43 59 55]
[ 34 36 191 44 32 34 50 39]
[ 46 52 147 59 43 180 21 54]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]]
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB set_trace >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/yiliu7/workspace/vllm/vllm/attention/ops/triton_unified_attention.py(264)kernel_unified_attention_2d()
-> S = tl.where(
(Pdb)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB continue >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
q_block_global_idx: [0], kv_head_idx: [1]
q_block_global_idx: [0], kv_head_idx: [1]
query_offset_0:
[0 0 0 0 1 1 1 1 2 2 2 2 3 3 3 3]
query_offset_1:
[4 5 6 7 4 5 6 7 4 5 6 7 4 5 6 7]
query_offset:
[[ 32 33 34 35 36 37 38 39]
[ 40 41 42 43 44 45 46 47]
[ 48 49 50 51 52 53 54 55]
[ 56 57 58 59 60 61 62 63]
[ 96 97 98 99 100 101 102 103]
[104 105 106 107 108 109 110 111]
[112 113 114 115 116 117 118 119]
[120 121 122 123 124 125 126 127]
[160 161 162 163 164 165 166 167]
[168 169 170 171 172 173 174 175]
[176 177 178 179 180 181 182 183]
[184 185 186 187 188 189 190 191]
[224 225 226 227 228 229 230 231]
[232 233 234 235 236 237 238 239]
[240 241 242 243 244 245 246 247]
[248 249 250 251 252 253 254 255]]
Q:
[[178 170 183 181 180 149 49 60]
[ 7 46 61 51 164 56 188 180]
[164 44 45 155 54 185 30 48]
[177 170 177 54 41 32 47 185]
[155 188 182 174 24 146 49 54]
[189 46 52 182 51 171 191 163]
[ 63 48 57 187 168 188 60 6]
[ 65 183 62 168 182 166 171 165]
[ 45 175 184 55 44 175 189 56]
[177 41 169 34 193 49 186 60]
[ 29 175 47 34 187 63 185 41]
[179 51 181 156 36 38 180 174]
[ 55 41 193 46 57 50 42 58]
[ 29 192 49 34 53 40 179 134]
[194 43 43 154 38 194 25 192]
[ 57 177 59 190 185 59 69 182]]
q_block_global_idx: [0], kv_head_idx: [1]
k_offset:
[[469256 469272 469288 469304 469320 469336 469352 469368 469384 469400
469416 469432 469448 469464 469480 469496]
[469257 469273 469289 469305 469321 469337 469353 469369 469385 469401
469417 469433 469449 469465 469481 469497]
[469258 469274 469290 469306 469322 469338 469354 469370 469386 469402
469418 469434 469450 469466 469482 469498]
[469259 469275 469291 469307 469323 469339 469355 469371 469387 469403
469419 469435 469451 469467 469483 469499]
[469260 469276 469292 469308 469324 469340 469356 469372 469388 469404
469420 469436 469452 469468 469484 469500]
[469261 469277 469293 469309 469325 469341 469357 469373 469389 469405
469421 469437 469453 469469 469485 469501]
[469262 469278 469294 469310 469326 469342 469358 469374 469390 469406
469422 469438 469454 469470 469486 469502]
[469263 469279 469295 469311 469327 469343 469359 469375 469391 469407
469423 469439 469455 469471 469487 469503]]
K_load:
[[ 57 42 50 60 67 52 0 0 0 0 0 0 0 0 0 0]
[ 54 157 164 52 42 179 0 0 0 0 0 0 0 0 0 0]
[ 58 176 52 152 41 57 0 0 0 0 0 0 0 0 0 0]
[ 51 184 189 182 43 50 0 0 0 0 0 0 0 0 0 0]
[ 49 40 51 41 192 169 0 0 0 0 0 0 0 0 0 0]
[134 57 64 173 64 168 0 0 0 0 0 0 0 0 0 0]
[181 176 46 191 152 59 0 0 0 0 0 0 0 0 0 0]
[ 63 181 191 178 186 40 0 0 0 0 0 0 0 0 0 0]]
v_offset:
[[469256 469257 469258 469259 469260 469261 469262 469263]
[469272 469273 469274 469275 469276 469277 469278 469279]
[469288 469289 469290 469291 469292 469293 469294 469295]
[469304 469305 469306 469307 469308 469309 469310 469311]
[469320 469321 469322 469323 469324 469325 469326 469327]
[469336 469337 469338 469339 469340 469341 469342 469343]
[469352 469353 469354 469355 469356 469357 469358 469359]
[469368 469369 469370 469371 469372 469373 469374 469375]
[469384 469385 469386 469387 469388 469389 469390 469391]
[469400 469401 469402 469403 469404 469405 469406 469407]
[469416 469417 469418 469419 469420 469421 469422 469423]
[469432 469433 469434 469435 469436 469437 469438 469439]
[469448 469449 469450 469451 469452 469453 469454 469455]
[469464 469465 469466 469467 469468 469469 469470 469471]
[469480 469481 469482 469483 469484 469485 469486 469487]
[469496 469497 469498 469499 469500 469501 469502 469503]]
V_load:
[[ 68 57 42 179 61 50 169 147]
[ 50 188 160 42 170 173 172 170]
[187 182 182 41 51 37 186 183]
[184 64 183 53 174 160 58 194]
[181 59 47 184 188 50 57 61]
[186 36 171 49 63 61 61 129]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]]
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB set_trace >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/yiliu7/workspace/vllm/vllm/attention/ops/triton_unified_attention.py(264)kernel_unified_attention_2d()
-> S = tl.where(
(Pdb)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB continue >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
q_block_global_idx: [1], kv_head_idx: [0]
q_block_global_idx: [1], kv_head_idx: [0]
query_offset_0:
[4 4 4 4 5 5 5 5 6 6 6 6 7 7 7 7]
query_offset_1:
[0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3]
query_offset:
[[256 257 258 259 260 261 262 263]
[264 265 266 267 268 269 270 271]
[272 273 274 275 276 277 278 279]
[280 281 282 283 284 285 286 287]
[320 321 322 323 324 325 326 327]
[328 329 330 331 332 333 334 335]
[336 337 338 339 340 341 342 343]
[344 345 346 347 348 349 350 351]
[384 385 386 387 388 389 390 391]
[392 393 394 395 396 397 398 399]
[400 401 402 403 404 405 406 407]
[408 409 410 411 412 413 414 415]
[448 449 450 451 452 453 454 455]
[456 457 458 459 460 461 462 463]
[464 465 466 467 468 469 470 471]
[472 473 474 475 476 477 478 479]]
Q:
[[ 39 57 168 176 185 59 55 151]
[ 45 175 28 57 51 164 175 190]
[ 25 184 162 50 45 175 55 181]
[145 186 178 57 179 57 179 161]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]]
q_block_global_idx: [1], kv_head_idx: [0]
k_offset:
[[469248 469264 469280 469296 469312 469328 469344 469360 469376 469392
469408 469424 469440 469456 469472 469488]
[469249 469265 469281 469297 469313 469329 469345 469361 469377 469393
469409 469425 469441 469457 469473 469489]
[469250 469266 469282 469298 469314 469330 469346 469362 469378 469394
469410 469426 469442 469458 469474 469490]
[469251 469267 469283 469299 469315 469331 469347 469363 469379 469395
469411 469427 469443 469459 469475 469491]
[469252 469268 469284 469300 469316 469332 469348 469364 469380 469396
469412 469428 469444 469460 469476 469492]
[469253 469269 469285 469301 469317 469333 469349 469365 469381 469397
469413 469429 469445 469461 469477 469493]
[469254 469270 469286 469302 469318 469334 469350 469366 469382 469398
469414 469430 469446 469462 469478 469494]
[469255 469271 469287 469303 469319 469335 469351 469367 469383 469399
469415 469431 469447 469463 469479 469495]]
K_load:
[[186 168 186 51 132 170 47 0 0 0 0 0 0 0 0 0]
[ 58 56 191 178 174 171 49 0 0 0 0 0 0 0 0 0]
[189 35 54 180 50 184 57 0 0 0 0 0 0 0 0 0]
[ 59 177 48 61 181 48 49 0 0 0 0 0 0 0 0 0]
[179 66 180 179 180 43 57 0 0 0 0 0 0 0 0 0]
[172 40 43 45 59 162 177 0 0 0 0 0 0 0 0 0]
[ 49 43 38 60 180 169 178 0 0 0 0 0 0 0 0 0]
[ 53 50 157 41 46 190 169 0 0 0 0 0 0 0 0 0]]
v_offset:
[[469248 469249 469250 469251 469252 469253 469254 469255]
[469264 469265 469266 469267 469268 469269 469270 469271]
[469280 469281 469282 469283 469284 469285 469286 469287]
[469296 469297 469298 469299 469300 469301 469302 469303]
[469312 469313 469314 469315 469316 469317 469318 469319]
[469328 469329 469330 469331 469332 469333 469334 469335]
[469344 469345 469346 469347 469348 469349 469350 469351]
[469360 469361 469362 469363 469364 469365 469366 469367]
[469376 469377 469378 469379 469380 469381 469382 469383]
[469392 469393 469394 469395 469396 469397 469398 469399]
[469408 469409 469410 469411 469412 469413 469414 469415]
[469424 469425 469426 469427 469428 469429 469430 469431]
[469440 469441 469442 469443 469444 469445 469446 469447]
[469456 469457 469458 469459 469460 469461 469462 469463]
[469472 469473 469474 469475 469476 469477 469478 469479]
[469488 469489 469490 469491 469492 469493 469494 469495]]
V_load:
[[181 50 178 57 177 187 59 21]
[173 57 39 41 42 194 183 54]
[ 17 179 181 53 151 162 60 186]
[177 170 39 179 188 43 59 55]
[ 34 36 191 44 32 34 50 39]
[ 46 52 147 59 43 180 21 54]
[ 1 193 157 154 183 175 43 61]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]]
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB set_trace >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/yiliu7/workspace/vllm/vllm/attention/ops/triton_unified_attention.py(264)kernel_unified_attention_2d()
-> S = tl.where(
(Pdb)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB continue >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
q_block_global_idx: [1], kv_head_idx: [1]
q_block_global_idx: [1], kv_head_idx: [1]
query_offset_0:
[4 4 4 4 5 5 5 5 6 6 6 6 7 7 7 7]
query_offset_1:
[4 5 6 7 4 5 6 7 4 5 6 7 4 5 6 7]
query_offset:
[[288 289 290 291 292 293 294 295]
[296 297 298 299 300 301 302 303]
[304 305 306 307 308 309 310 311]
[312 313 314 315 316 317 318 319]
[352 353 354 355 356 357 358 359]
[360 361 362 363 364 365 366 367]
[368 369 370 371 372 373 374 375]
[376 377 378 379 380 381 382 383]
[416 417 418 419 420 421 422 423]
[424 425 426 427 428 429 430 431]
[432 433 434 435 436 437 438 439]
[440 441 442 443 444 445 446 447]
[480 481 482 483 484 485 486 487]
[488 489 490 491 492 493 494 495]
[496 497 498 499 500 501 502 503]
[504 505 506 507 508 509 510 511]]
Q:
[[ 51 179 36 190 177 179 160 178]
[165 62 49 152 64 168 190 174]
[185 58 43 179 174 184 183 49]
[ 58 191 55 180 19 63 181 179]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]]
q_block_global_idx: [1], kv_head_idx: [1]
k_offset:
[[469256 469272 469288 469304 469320 469336 469352 469368 469384 469400
469416 469432 469448 469464 469480 469496]
[469257 469273 469289 469305 469321 469337 469353 469369 469385 469401
469417 469433 469449 469465 469481 469497]
[469258 469274 469290 469306 469322 469338 469354 469370 469386 469402
469418 469434 469450 469466 469482 469498]
[469259 469275 469291 469307 469323 469339 469355 469371 469387 469403
469419 469435 469451 469467 469483 469499]
[469260 469276 469292 469308 469324 469340 469356 469372 469388 469404
469420 469436 469452 469468 469484 469500]
[469261 469277 469293 469309 469325 469341 469357 469373 469389 469405
469421 469437 469453 469469 469485 469501]
[469262 469278 469294 469310 469326 469342 469358 469374 469390 469406
469422 469438 469454 469470 469486 469502]
[469263 469279 469295 469311 469327 469343 469359 469375 469391 469407
469423 469439 469455 469471 469487 469503]]
K_load:
[[ 57 42 50 60 67 52 183 0 0 0 0 0 0 0 0 0]
[ 54 157 164 52 42 179 189 0 0 0 0 0 0 0 0 0]
[ 58 176 52 152 41 57 186 0 0 0 0 0 0 0 0 0]
[ 51 184 189 182 43 50 164 0 0 0 0 0 0 0 0 0]
[ 49 40 51 41 192 169 51 0 0 0 0 0 0 0 0 0]
[134 57 64 173 64 168 33 0 0 0 0 0 0 0 0 0]
[181 176 46 191 152 59 184 0 0 0 0 0 0 0 0 0]
[ 63 181 191 178 186 40 48 0 0 0 0 0 0 0 0 0]]
v_offset:
[[469256 469257 469258 469259 469260 469261 469262 469263]
[469272 469273 469274 469275 469276 469277 469278 469279]
[469288 469289 469290 469291 469292 469293 469294 469295]
[469304 469305 469306 469307 469308 469309 469310 469311]
[469320 469321 469322 469323 469324 469325 469326 469327]
[469336 469337 469338 469339 469340 469341 469342 469343]
[469352 469353 469354 469355 469356 469357 469358 469359]
[469368 469369 469370 469371 469372 469373 469374 469375]
[469384 469385 469386 469387 469388 469389 469390 469391]
[469400 469401 469402 469403 469404 469405 469406 469407]
[469416 469417 469418 469419 469420 469421 469422 469423]
[469432 469433 469434 469435 469436 469437 469438 469439]
[469448 469449 469450 469451 469452 469453 469454 469455]
[469464 469465 469466 469467 469468 469469 469470 469471]
[469480 469481 469482 469483 469484 469485 469486 469487]
[469496 469497 469498 469499 469500 469501 469502 469503]]
V_load:
[[ 68 57 42 179 61 50 169 147]
[ 50 188 160 42 170 173 172 170]
[187 182 182 41 51 37 186 183]
[184 64 183 53 174 160 58 194]
[181 59 47 184 188 50 57 61]
[186 36 171 49 63 61 61 129]
[ 55 54 173 56 63 60 45 56]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0]]
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB set_trace >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/yiliu7/workspace/vllm/vllm/attention/ops/triton_unified_attention.py(264)kernel_unified_attention_2d()
-> S = tl.where(
(Pdb)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB continue >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB set_trace >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/yiliu7/workspace/vllm/tests/kernels/attention/test_triton_unified_attention.py(256)test_triton_unified_attn()
-> torch.testing.assert_close(output, ref_output, atol=atol, rtol=rtol),
(Pdb)
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB continue >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
PASSED
=============================== warnings summary ===============================
<frozen importlib._bootstrap>:488
<frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyPacked has no __module__ attribute
<frozen importlib._bootstrap>:488
<frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyObject has no __module__ attribute
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
/usr/local/lib/python3.12/dist-packages/triton/runtime/interpreter.py:818: DeprecationWarning: Conversion of an array with ndim > 0 to a scalar is deprecated, and will error in future. Ensure you extract a single element from your array before performing this operation. (Deprecated NumPy 1.25.)
tensor.__index__ = lambda self: int(self.handle.data)
tests/kernels/attention/test_triton_unified_attention.py::test_triton_unified_attn[q_dtype0-2048-None-dtype0-None-16-8-num_heads0-seq_lens0]
/usr/local/lib/python3.12/dist-packages/triton/runtime/interpreter.py:463: RuntimeWarning: invalid value encountered in divide
return TensorHandle(op(lhs.data, rhs.data), lhs.dtype.scalar)
-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
======================= 1 passed, 11 warnings in 10.85s ========================
sys:1: DeprecationWarning: builtin type swigvarlink has no __module__ attribute