@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 9d7577b2bd
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 8, 00:00 UTC

[Kernel] [Helion] [9/N] Canonicalize GPU variant names to base model names (#34928)

Signed-off-by: Yanan Cao <gmagogsfm@gmail.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Yanan Cao committed Feb 20, 2026 at 19:55 UTC 9d7577b2bdf9de1155f1078e591446453723a88c
4 files changed +59 -27733
tests/kernels/helion/test_utils.py
+5 -3
@@ -11,11 +11,13 @@ from vllm.kernels.helion.utils import canonicalize_gpu_name
11 "driver_reported_name,expected",
12 [
13 ("NVIDIA H200", "nvidia_h200"),
14 - ("NVIDIA A100-SXM4-80GB", "nvidia_a100_sxm4_80gb"),
15 - ("NVIDIA H100 80GB HBM3", "nvidia_h100_80gb_hbm3"),
14 + ("NVIDIA A100-SXM4-80GB", "nvidia_a100"),
15 + ("NVIDIA H100 80GB HBM3", "nvidia_h100"),
16 + ("NVIDIA H100 PCIe", "nvidia_h100"),
17 + ("NVIDIA H100 SXM5", "nvidia_h100"),
18 ("NVIDIA GeForce RTX 4090", "nvidia_geforce_rtx_4090"),
19 ("AMD Instinct MI300X", "amd_instinct_mi300x"),
18 - ("Tesla V100-SXM2-32GB", "tesla_v100_sxm2_32gb"),
20 + ("Tesla V100-SXM2-32GB", "tesla_v100"),
21 ],
22 )
23 def test_canonicalize_gpu_name(driver_reported_name, expected):
vllm/kernels/helion/config_manager.py
+9 -1
@@ -71,10 +71,18 @@ class ConfigSet:
71 platform_dict = self._configs.get(platform)
72 if platform_dict is None:
73 avail_platforms = self.get_platforms()
74 + # TODO(@gmagogsfm): add a CLI/env override flag so users can
75 + # directly specify a platform name instead of relying on
76 + # auto-detection, and suggest it in this error message.
77 raise KeyError(
78 f"Config not found for kernel '{self._kernel_name}': "
79 f"platform '{platform}' not found. "
77 - f"Available platforms: {avail_platforms or '(none)'}"
80 + f"Available platforms: {avail_platforms or '(none)'}. "
81 + f"If your GPU is a variant of a supported platform, "
82 + f"consider adding a mapping in _GPU_NAME_ALIASES in "
83 + f"vllm/kernels/helion/utils.py, or run "
84 + f"scripts/autotune_helion_kernels.py to generate configs "
85 + f"for your platform."
86 )
87
88 config = platform_dict.get(config_key)
vllm/kernels/helion/configs/silu_mul_fp8.json
-27724
@@ -13861,27730 +13861,6 @@
13861 "range_warp_specializes": []
13862 }
13863 },
13864 - "nvidia_h100_pcie": {
13865 - "intermediate_2048_numtokens_256": {
13866 - "block_sizes": [
13867 - 64,
13868 - 32
13869 - ],
13870 - "loop_orders": [
13871 - [
13872 - 0,
13873 - 1
13874 - ]
13875 - ],
13876 - "flatten_loops": [
13877 - true
13878 - ],
13879 - "l2_groupings": [
13880 - 1
13881 - ],
13882 - "range_unroll_factors": [
13883 - 0
13884 - ],
13885 - "range_warp_specializes": [],
13886 - "range_num_stages": [
13887 - 0
13888 - ],
13889 - "range_multi_buffers": [
13890 - null
13891 - ],
13892 - "range_flattens": [
13893 - null
13894 - ],
13895 - "load_eviction_policies": [
13896 - "",
13897 - "",
13898 - ""
13899 - ],
13900 - "num_warps": 8,
13901 - "num_stages": 1,
13902 - "indexing": [
13903 - "pointer",
13904 - "pointer",
13905 - "pointer",
13906 - "pointer"
13907 - ],
13908 - "pid_type": "flat"
13909 - },
13910 - "intermediate_4096_numtokens_256": {
13911 - "block_sizes": [
13912 - 32,
13913 - 512
13914 - ],
13915 - "loop_orders": [
13916 - [
13917 - 0,
13918 - 1
13919 - ]
13920 - ],
13921 - "flatten_loops": [
13922 - true
13923 - ],
13924 - "l2_groupings": [
13925 - 2
13926 - ],
13927 - "range_unroll_factors": [
13928 - 0
13929 - ],
13930 - "range_warp_specializes": [],
13931 - "range_num_stages": [
13932 - 0
13933 - ],
13934 - "range_multi_buffers": [
13935 - null
13936 - ],
13937 - "range_flattens": [
13938 - null
13939 - ],
13940 - "load_eviction_policies": [
13941 - "",
13942 - "",
13943 - ""
13944 - ],
13945 - "num_warps": 2,
13946 - "num_stages": 2,
13947 - "indexing": [
13948 - "pointer",
13949 - "tensor_descriptor",
13950 - "pointer",
13951 - "pointer"
13952 - ],
13953 - "pid_type": "flat"
13954 - },
13955 - "default": {
13956 - "block_sizes": [
13957 - 1,
13958 - 512
13959 - ],
13960 - "loop_orders": [
13961 - [
13962 - 1,
13963 - 0
13964 - ]
13965 - ],
13966 - "flatten_loops": [
13967 - false
13968 - ],
13969 - "l2_groupings": [
13970 - 4
13971 - ],
13972 - "range_unroll_factors": [
13973 - 0
13974 - ],
13975 - "range_warp_specializes": [],
13976 - "range_num_stages": [
13977 - 0
13978 - ],
13979 - "range_multi_buffers": [
13980 - null
13981 - ],
13982 - "range_flattens": [
13983 - null
13984 - ],
13985 - "load_eviction_policies": [
13986 - "",
13987 - "first",
13988 - ""
13989 - ],
13990 - "num_warps": 8,
13991 - "num_stages": 2,
13992 - "indexing": [
13993 - "tensor_descriptor",
13994 - "tensor_descriptor",
13995 - "tensor_descriptor",
13996 - "pointer"
13997 - ],
13998 - "pid_type": "flat"
13999 - },
14000 - "intermediate_8192_numtokens_256": {
14001 - "block_sizes": [
14002 - 32,
14003 - 8
14004 - ],
14005 - "loop_orders": [
14006 - [
14007 - 0,
14008 - 1
14009 - ]
14010 - ],
14011 - "flatten_loops": [
14012 - true
14013 - ],
14014 - "l2_groupings": [
14015 - 1
14016 - ],
14017 - "range_unroll_factors": [
14018 - 0
14019 - ],
14020 - "range_warp_specializes": [],
14021 - "range_num_stages": [
14022 - 0
14023 - ],
14024 - "range_multi_buffers": [
14025 - null
14026 - ],
14027 - "range_flattens": [
14028 - null
14029 - ],
14030 - "load_eviction_policies": [
14031 - "",
14032 - "first",
14033 - ""
14034 - ],
14035 - "num_warps": 4,
14036 - "num_stages": 1,
14037 - "indexing": [
14038 - "tensor_descriptor",
14039 - "pointer",
14040 - "pointer",
14041 - "tensor_descriptor"
14042 - ],
14043 - "pid_type": "flat"
14044 - },
14045 - "intermediate_14336_numtokens_256": {
14046 - "block_sizes": [
14047 - 16,
14048 - 32
14049 - ],
14050 - "loop_orders": [
14051 - [
14052 - 0,
14053 - 1
14054 - ]
14055 - ],
14056 - "flatten_loops": [
14057 - true
14058 - ],
14059 - "l2_groupings": [
14060 - 1
14061 - ],
14062 - "range_unroll_factors": [
14063 - 0
14064 - ],
14065 - "range_warp_specializes": [],
14066 - "range_num_stages": [
14067 - 0
14068 - ],
14069 - "range_multi_buffers": [
14070 - null
14071 - ],
14072 - "range_flattens": [
14073 - null
14074 - ],
14075 - "load_eviction_policies": [
14076 - "",
14077 - "",
14078 - ""
14079 - ],
14080 - "num_warps": 1,
14081 - "num_stages": 1,
14082 - "indexing": [
14083 - "pointer",
14084 - "pointer",
14085 - "pointer",
14086 - "tensor_descriptor"
14087 - ],
14088 - "pid_type": "flat"
14089 - },
14090 - "intermediate_11008_numtokens_256": {
14091 - "block_sizes": [
14092 - 64,
14093 - 32
14094 - ],
14095 - "loop_orders": [
14096 - [
14097 - 0,
14098 - 1
14099 - ]
14100 - ],
14101 - "flatten_loops": [
14102 - true
14103 - ],
14104 - "l2_groupings": [
14105 - 1
14106 - ],
14107 - "range_unroll_factors": [
14108 - 0
14109 - ],
14110 - "range_warp_specializes": [],
14111 - "range_num_stages": [
14112 - 0
14113 - ],
14114 - "range_multi_buffers": [
14115 - null
14116 - ],
14117 - "range_flattens": [
14118 - null
14119 - ],
14120 - "load_eviction_policies": [
14121 - "",
14122 - "",
14123 - ""
14124 - ],
14125 - "num_warps": 2,
14126 - "num_stages": 1,
14127 - "indexing": [
14128 - "pointer",
14129 - "pointer",
14130 - "pointer",
14131 - "pointer"
14132 - ],
14133 - "pid_type": "flat"
14134 - },
14135 - "intermediate_7688_numtokens_256": {
14136 - "block_sizes": [
14137 - 8,
14138 - 16
14139 - ],
14140 - "loop_orders": [
14141 - [
14142 - 0,
14143 - 1
14144 - ]
14145 - ],
14146 - "flatten_loops": [
14147 - true
14148 - ],
14149 - "l2_groupings": [
14150 - 1
14151 - ],
14152 - "range_unroll_factors": [
14153 - 0
14154 - ],
14155 - "range_warp_specializes": [],
14156 - "range_num_stages": [
14157 - 0
14158 - ],
14159 - "range_multi_buffers": [
14160 - null
14161 - ],
14162 - "range_flattens": [
14163 - null
14164 - ],
14165 - "load_eviction_policies": [
14166 - "",
14167 - "last",
14168 - ""
14169 - ],
14170 - "num_warps": 1,
14171 - "num_stages": 1,
14172 - "indexing": [
14173 - "pointer",
14174 - "pointer",
14175 - "pointer",
14176 - "pointer"
14177 - ],
14178 - "pid_type": "flat"
14179 - },
14180 - "intermediate_2880_numtokens_256": {
14181 - "block_sizes": [
14182 - 32,
14183 - 64
14184 - ],
14185 - "loop_orders": [
14186 - [
14187 - 0,
14188 - 1
14189 - ]
14190 - ],
14191 - "flatten_loops": [
14192 - true
14193 - ],
14194 - "l2_groupings": [
14195 - 2
14196 - ],
14197 - "range_unroll_factors": [
14198 - 0
14199 - ],
14200 - "range_warp_specializes": [],
14201 - "range_num_stages": [
14202 - 0
14203 - ],
14204 - "range_multi_buffers": [
14205 - null
14206 - ],
14207 - "range_flattens": [
14208 - null
14209 - ],
14210 - "load_eviction_policies": [
14211 - "",
14212 - "",
14213 - ""
14214 - ],
14215 - "num_warps": 8,
14216 - "num_stages": 1,
14217 - "indexing": [
14218 - "pointer",
14219 - "pointer",
14220 - "pointer",
14221 - "pointer"
14222 - ],
14223 - "pid_type": "flat"
14224 - },
14225 - "intermediate_2048_numtokens_1": {
14226 - "block_sizes": [
14227 - 1,
14228 - 16
14229 - ],
14230 - "loop_orders": [
14231 - [
14232 - 0,
14233 - 1
14234 - ]
14235 - ],
14236 - "flatten_loops": [
14237 - false
14238 - ],
14239 - "l2_groupings": [
14240 - 1
14241 - ],
14242 - "range_unroll_factors": [
14243 - 0
14244 - ],
14245 - "range_warp_specializes": [],
14246 - "range_num_stages": [
14247 - 0
14248 - ],
14249 - "range_multi_buffers": [
14250 - null
14251 - ],
14252 - "range_flattens": [
14253 - null
14254 - ],
14255 - "load_eviction_policies": [
14256 - "",
14257 - "",
14258 - ""
14259 - ],
14260 - "num_warps": 16,
14261 - "num_stages": 2,
14262 - "indexing": [
14263 - "pointer",
14264 - "pointer",
14265 - "pointer",
14266 - "tensor_descriptor"
14267 - ],
14268 - "pid_type": "flat"
14269 - },
14270 - "intermediate_2880_numtokens_1": {
14271 - "block_sizes": [
14272 - 1,
14273 - 1
14274 - ],
14275 - "loop_orders": [
14276 - [
14277 - 0,
14278 - 1
14279 - ]
14280 - ],
14281 - "flatten_loops": [
14282 - false
14283 - ],
14284 - "l2_groupings": [
14285 - 1
14286 - ],
14287 - "range_unroll_factors": [
14288 - 0
14289 - ],
14290 - "range_warp_specializes": [],
14291 - "range_num_stages": [
14292 - 0
14293 - ],
14294 - "range_multi_buffers": [
14295 - null
14296 - ],
14297 - "range_flattens": [
14298 - null
14299 - ],
14300 - "load_eviction_policies": [
14301 - "",
14302 - "",
14303 - ""
14304 - ],
14305 - "num_warps": 16,
14306 - "num_stages": 2,
14307 - "indexing": [
14308 - "pointer",
14309 - "pointer",
14310 - "pointer",
14311 - "pointer"
14312 - ],
14313 - "pid_type": "flat"
14314 - },
14315 - "intermediate_4096_numtokens_1": {
14316 - "block_sizes": [
14317 - 1,
14318 - 32
14319 - ],
14320 - "loop_orders": [
14321 - [
14322 - 1,
14323 - 0
14324 - ]
14325 - ],
14326 - "flatten_loops": [
14327 - false
14328 - ],
14329 - "l2_groupings": [
14330 - 1
14331 - ],
14332 - "range_unroll_factors": [
14333 - 0
14334 - ],
14335 - "range_warp_specializes": [],
14336 - "range_num_stages": [
14337 - 0
14338 - ],
14339 - "range_multi_buffers": [
14340 - null
14341 - ],
14342 - "range_flattens": [
14343 - null
14344 - ],
14345 - "load_eviction_policies": [
14346 - "",
14347 - "",
14348 - ""
14349 - ],
14350 - "num_warps": 4,
14351 - "num_stages": 2,
14352 - "indexing": [
14353 - "pointer",
14354 - "pointer",
14355 - "pointer",
14356 - "pointer"
14357 - ],
14358 - "pid_type": "flat"
14359 - },
14360 - "intermediate_8192_numtokens_1": {
14361 - "block_sizes": [
14362 - 1,
14363 - 32
14364 - ],
14365 - "loop_orders": [
14366 - [
14367 - 0,
14368 - 1
14369 - ]
14370 - ],
14371 - "flatten_loops": [
14372 - false
14373 - ],
14374 - "l2_groupings": [
14375 - 1
14376 - ],
14377 - "range_unroll_factors": [
14378 - 0
14379 - ],
14380 - "range_warp_specializes": [],
14381 - "range_num_stages": [
14382 - 0
14383 - ],
14384 - "range_multi_buffers": [
14385 - null
14386 - ],
14387 - "range_flattens": [
14388 - null
14389 - ],
14390 - "load_eviction_policies": [
14391 - "",
14392 - "",
14393 - ""
14394 - ],
14395 - "num_warps": 1,
14396 - "num_stages": 2,
14397 - "indexing": [
14398 - "pointer",
14399 - "pointer",
14400 - "pointer",
14401 - "pointer"
14402 - ],
14403 - "pid_type": "flat"
14404 - },
14405 - "intermediate_11008_numtokens_1": {
14406 - "block_sizes": [
14407 - 1,
14408 - 32
14409 - ],
14410 - "loop_orders": [
14411 - [
14412 - 0,
14413 - 1
14414 - ]
14415 - ],
14416 - "flatten_loops": [
14417 - true
14418 - ],
14419 - "l2_groupings": [
14420 - 1
14421 - ],
14422 - "range_unroll_factors": [
14423 - 0
14424 - ],
14425 - "range_warp_specializes": [],
14426 - "range_num_stages": [
14427 - 0
14428 - ],
14429 - "range_multi_buffers": [
14430 - null
14431 - ],
14432 - "range_flattens": [
14433 - null
14434 - ],
14435 - "load_eviction_policies": [
14436 - "first",
14437 - "",
14438 - ""
14439 - ],
14440 - "num_warps": 4,
14441 - "num_stages": 1,
14442 - "indexing": [
14443 - "pointer",
14444 - "pointer",
14445 - "pointer",
14446 - "pointer"
14447 - ],
14448 - "pid_type": "flat"
14449 - },
14450 - "intermediate_14336_numtokens_1": {
14451 - "block_sizes": [
14452 - 1,
14453 - 32
14454 - ],
14455 - "loop_orders": [
14456 - [
14457 - 0,
14458 - 1
14459 - ]
14460 - ],
14461 - "flatten_loops": [
14462 - false
14463 - ],
14464 - "l2_groupings": [
14465 - 2
14466 - ],
14467 - "range_unroll_factors": [
14468 - 0
14469 - ],
14470 - "range_warp_specializes": [],
14471 - "range_num_stages": [
14472 - 0
14473 - ],
14474 - "range_multi_buffers": [
14475 - null
14476 - ],
14477 - "range_flattens": [
14478 - null
14479 - ],
14480 - "load_eviction_policies": [
14481 - "",
14482 - "",
14483 - "first"
14484 - ],
14485 - "num_warps": 8,
14486 - "num_stages": 1,
14487 - "indexing": [
14488 - "pointer",
14489 - "pointer",
14490 - "pointer",
14491 - "pointer"
14492 - ],
14493 - "pid_type": "flat"
14494 - },
14495 - "intermediate_2048_numtokens_2": {
14496 - "block_sizes": [
14497 - 2,
14498 - 16
14499 - ],
14500 - "loop_orders": [
14501 - [
14502 - 0,
14503 - 1
14504 - ]
14505 - ],
14506 - "flatten_loops": [
14507 - true
14508 - ],
14509 - "l2_groupings": [
14510 - 2
14511 - ],
14512 - "range_unroll_factors": [
14513 - 0
14514 - ],
14515 - "range_warp_specializes": [],
14516 - "range_num_stages": [
14517 - 0
14518 - ],
14519 - "range_multi_buffers": [
14520 - null
14521 - ],
14522 - "range_flattens": [
14523 - null
14524 - ],
14525 - "load_eviction_policies": [
14526 - "",
14527 - "",
14528 - ""
14529 - ],
14530 - "num_warps": 2,
14531 - "num_stages": 1,
14532 - "indexing": [
14533 - "pointer",
14534 - "pointer",
14535 - "tensor_descriptor",
14536 - "pointer"
14537 - ],
14538 - "pid_type": "flat"
14539 - },
14540 - "intermediate_2880_numtokens_2": {
14541 - "block_sizes": [
14542 - 1,
14543 - 4
14544 - ],
14545 - "loop_orders": [
14546 - [
14547 - 0,
14548 - 1
14549 - ]
14550 - ],
14551 - "flatten_loops": [
14552 - false
14553 - ],
14554 - "l2_groupings": [
14555 - 1
14556 - ],
14557 - "range_unroll_factors": [
14558 - 0
14559 - ],
14560 - "range_warp_specializes": [],
14561 - "range_num_stages": [
14562 - 0
14563 - ],
14564 - "range_multi_buffers": [
14565 - null
14566 - ],
14567 - "range_flattens": [
14568 - null
14569 - ],
14570 - "load_eviction_policies": [
14571 - "first",
14572 - "",
14573 - ""
14574 - ],
14575 - "num_warps": 32,
14576 - "num_stages": 1,
14577 - "indexing": [
14578 - "pointer",
14579 - "pointer",
14580 - "pointer",
14581 - "pointer"
14582 - ],
14583 - "pid_type": "flat"
14584 - },
14585 - "intermediate_4096_numtokens_2": {
14586 - "block_sizes": [
14587 - 2,
14588 - 32
14589 - ],
14590 - "loop_orders": [
14591 - [
14592 - 0,
14593 - 1
14594 - ]
14595 - ],
14596 - "flatten_loops": [
14597 - true
14598 - ],
14599 - "l2_groupings": [
14600 - 1
14601 - ],
14602 - "range_unroll_factors": [
14603 - 0
14604 - ],
14605 - "range_warp_specializes": [],
14606 - "range_num_stages": [
14607 - 0
14608 - ],
14609 - "range_multi_buffers": [
14610 - null
14611 - ],
14612 - "range_flattens": [
14613 - null
14614 - ],
14615 - "load_eviction_policies": [
14616 - "",
14617 - "",
14618 - ""
14619 - ],
14620 - "num_warps": 8,
14621 - "num_stages": 2,
14622 - "indexing": [
14623 - "pointer",
14624 - "pointer",
14625 - "pointer",
14626 - "pointer"
14627 - ],
14628 - "pid_type": "flat"
14629 - },
14630 - "intermediate_8192_numtokens_2": {
14631 - "block_sizes": [
14632 - 1,
14633 - 32
14634 - ],
14635 - "loop_orders": [
14636 - [
14637 - 0,
14638 - 1
14639 - ]
14640 - ],
14641 - "flatten_loops": [
14642 - false
14643 - ],
14644 - "l2_groupings": [
14645 - 1
14646 - ],
14647 - "range_unroll_factors": [
14648 - 0
14649 - ],
14650 - "range_warp_specializes": [],
14651 - "range_num_stages": [
14652 - 0
14653 - ],
14654 - "range_multi_buffers": [
14655 - null
14656 - ],
14657 - "range_flattens": [
14658 - null
14659 - ],
14660 - "load_eviction_policies": [
14661 - "first",
14662 - "",
14663 - ""
14664 - ],
14665 - "num_warps": 8,
14666 - "num_stages": 2,
14667 - "indexing": [
14668 - "tensor_descriptor",
14669 - "pointer",
14670 - "pointer",
14671 - "pointer"
14672 - ],
14673 - "pid_type": "flat"
14674 - },
14675 - "intermediate_11008_numtokens_2": {
14676 - "block_sizes": [
14677 - 1,
14678 - 256
14679 - ],
14680 - "loop_orders": [
14681 - [
14682 - 0,
14683 - 1
14684 - ]
14685 - ],
14686 - "flatten_loops": [
14687 - false
14688 - ],
14689 - "l2_groupings": [
14690 - 1
14691 - ],
14692 - "range_unroll_factors": [
14693 - 0
14694 - ],
14695 - "range_warp_specializes": [],
14696 - "range_num_stages": [
14697 - 0
14698 - ],
14699 - "range_multi_buffers": [
14700 - null
14701 - ],
14702 - "range_flattens": [
14703 - null
14704 - ],
14705 - "load_eviction_policies": [
14706 - "",
14707 - "first",
14708 - ""
14709 - ],
14710 - "num_warps": 4,
14711 - "num_stages": 2,
14712 - "indexing": [
14713 - "pointer",
14714 - "pointer",
14715 - "pointer",
14716 - "tensor_descriptor"
14717 - ],
14718 - "pid_type": "flat"
14719 - },
14720 - "intermediate_14336_numtokens_2": {
14721 - "block_sizes": [
14722 - 1,
14723 - 64
14724 - ],
14725 - "loop_orders": [
14726 - [
14727 - 0,
14728 - 1
14729 - ]
14730 - ],
14731 - "flatten_loops": [
14732 - false
14733 - ],
14734 - "l2_groupings": [
14735 - 1
14736 - ],
14737 - "range_unroll_factors": [
14738 - 0
14739 - ],
14740 - "range_warp_specializes": [],
14741 - "range_num_stages": [
14742 - 0
14743 - ],
14744 - "range_multi_buffers": [
14745 - null
14746 - ],
14747 - "range_flattens": [
14748 - null
14749 - ],
14750 - "load_eviction_policies": [
14751 - "",
14752 - "",
14753 - ""
14754 - ],
14755 - "num_warps": 8,
14756 - "num_stages": 2,
14757 - "indexing": [
14758 - "pointer",
14759 - "pointer",
14760 - "pointer",
14761 - "pointer"
14762 - ],
14763 - "pid_type": "flat"
14764 - },
14765 - "intermediate_2048_numtokens_4": {
14766 - "block_sizes": [
14767 - 1,
14768 - 256
14769 - ],
14770 - "loop_orders": [
14771 - [
14772 - 0,
14773 - 1
14774 - ]
14775 - ],
14776 - "flatten_loops": [
14777 - true
14778 - ],
14779 - "l2_groupings": [
14780 - 1
14781 - ],
14782 - "range_unroll_factors": [
14783 - 0
14784 - ],
14785 - "range_warp_specializes": [],
14786 - "range_num_stages": [
14787 - 0
14788 - ],
14789 - "range_multi_buffers": [
14790 - null
14791 - ],
14792 - "range_flattens": [
14793 - null
14794 - ],
14795 - "load_eviction_policies": [
14796 - "",
14797 - "",
14798 - ""
14799 - ],
14800 - "num_warps": 32,
14801 - "num_stages": 2,
14802 - "indexing": [
14803 - "pointer",
14804 - "pointer",
14805 - "pointer",
14806 - "pointer"
14807 - ],
14808 - "pid_type": "flat"
14809 - },
14810 - "intermediate_2880_numtokens_4": {
14811 - "block_sizes": [
14812 - 1,
14813 - 8
14814 - ],
14815 - "loop_orders": [
14816 - [
14817 - 0,
14818 - 1
14819 - ]
14820 - ],
14821 - "flatten_loops": [
14822 - true
14823 - ],
14824 - "l2_groupings": [
14825 - 1
14826 - ],
14827 - "range_unroll_factors": [
14828 - 0
14829 - ],
14830 - "range_warp_specializes": [],
14831 - "range_num_stages": [
14832 - 0
14833 - ],
14834 - "range_multi_buffers": [
14835 - null
14836 - ],
14837 - "range_flattens": [
14838 - null
14839 - ],
14840 - "load_eviction_policies": [
14841 - "",
14842 - "",
14843 - ""
14844 - ],
14845 - "num_warps": 1,
14846 - "num_stages": 2,
14847 - "indexing": [
14848 - "pointer",
14849 - "pointer",
14850 - "pointer",
14851 - "pointer"
14852 - ],
14853 - "pid_type": "flat"
14854 - },
14855 - "intermediate_4096_numtokens_4": {
14856 - "block_sizes": [
14857 - 4,
14858 - 16
14859 - ],
14860 - "loop_orders": [
14861 - [
14862 - 0,
14863 - 1
14864 - ]
14865 - ],
14866 - "flatten_loops": [
14867 - true
14868 - ],
14869 - "l2_groupings": [
14870 - 2
14871 - ],
14872 - "range_unroll_factors": [
14873 - 0
14874 - ],
14875 - "range_warp_specializes": [],
14876 - "range_num_stages": [
14877 - 0
14878 - ],
14879 - "range_multi_buffers": [
14880 - null
14881 - ],
14882 - "range_flattens": [
14883 - null
14884 - ],
14885 - "load_eviction_policies": [
14886 - "last",
14887 - "",
14888 - ""
14889 - ],
14890 - "num_warps": 4,
14891 - "num_stages": 1,
14892 - "indexing": [
14893 - "pointer",
14894 - "pointer",
14895 - "pointer",
14896 - "pointer"
14897 - ],
14898 - "pid_type": "flat"
14899 - },
14900 - "intermediate_8192_numtokens_4": {
14901 - "block_sizes": [
14902 - 1,
14903 - 16
14904 - ],
14905 - "loop_orders": [
14906 - [
14907 - 0,
14908 - 1
14909 - ]
14910 - ],
14911 - "flatten_loops": [
14912 - true
14913 - ],
14914 - "l2_groupings": [
14915 - 1
14916 - ],
14917 - "range_unroll_factors": [
14918 - 0
14919 - ],
14920 - "range_warp_specializes": [],
14921 - "range_num_stages": [
14922 - 0
14923 - ],
14924 - "range_multi_buffers": [
14925 - null
14926 - ],
14927 - "range_flattens": [
14928 - null
14929 - ],
14930 - "load_eviction_policies": [
14931 - "",
14932 - "",
14933 - ""
14934 - ],
14935 - "num_warps": 4,
14936 - "num_stages": 1,
14937 - "indexing": [
14938 - "pointer",
14939 - "pointer",
14940 - "pointer",
14941 - "pointer"
14942 - ],
14943 - "pid_type": "flat"
14944 - },
14945 - "intermediate_11008_numtokens_4": {
14946 - "block_sizes": [
14947 - 1,
14948 - 32
14949 - ],
14950 - "loop_orders": [
14951 - [
14952 - 0,
14953 - 1
14954 - ]
14955 - ],
14956 - "flatten_loops": [
14957 - false
14958 - ],
14959 - "l2_groupings": [
14960 - 1
14961 - ],
14962 - "range_unroll_factors": [
14963 - 0
14964 - ],
14965 - "range_warp_specializes": [],
14966 - "range_num_stages": [
14967 - 0
14968 - ],
14969 - "range_multi_buffers": [
14970 - null
14971 - ],
14972 - "range_flattens": [
14973 - null
14974 - ],
14975 - "load_eviction_policies": [
14976 - "",
14977 - "first",
14978 - ""
14979 - ],
14980 - "num_warps": 16,
14981 - "num_stages": 1,
14982 - "indexing": [
14983 - "pointer",
14984 - "pointer",
14985 - "tensor_descriptor",
14986 - "pointer"
14987 - ],
14988 - "pid_type": "flat"
14989 - },
14990 - "intermediate_14336_numtokens_4": {
14991 - "block_sizes": [
14992 - 4,
14993 - 16
14994 - ],
14995 - "loop_orders": [
14996 - [
14997 - 1,
14998 - 0
14999 - ]
15000 - ],
15001 - "flatten_loops": [
15002 - true
15003 - ],
15004 - "l2_groupings": [
15005 - 1
15006 - ],
15007 - "range_unroll_factors": [
15008 - 0
15009 - ],
15010 - "range_warp_specializes": [],
15011 - "range_num_stages": [
15012 - 0
15013 - ],
15014 - "range_multi_buffers": [
15015 - null
15016 - ],
15017 - "range_flattens": [
15018 - null
15019 - ],
15020 - "load_eviction_policies": [
15021 - "",
15022 - "",
15023 - ""
15024 - ],
15025 - "num_warps": 16,
15026 - "num_stages": 1,
15027 - "indexing": [
15028 - "pointer",
15029 - "pointer",
15030 - "pointer",
15031 - "pointer"
15032 - ],
15033 - "pid_type": "flat"
15034 - },
15035 - "intermediate_2048_numtokens_8": {
15036 - "block_sizes": [
15037 - 8,
15038 - 256
15039 - ],
15040 - "loop_orders": [
15041 - [
15042 - 1,
15043 - 0
15044 - ]
15045 - ],
15046 - "flatten_loops": [
15047 - true
15048 - ],
15049 - "l2_groupings": [
15050 - 1
15051 - ],
15052 - "range_unroll_factors": [
15053 - 0
15054 - ],
15055 - "range_warp_specializes": [],
15056 - "range_num_stages": [
15057 - 0
15058 - ],
15059 - "range_multi_buffers": [
15060 - null
15061 - ],
15062 - "range_flattens": [
15063 - null
15064 - ],
15065 - "load_eviction_policies": [
15066 - "last",
15067 - "",
15068 - ""
15069 - ],
15070 - "num_warps": 32,
15071 - "num_stages": 1,
15072 - "indexing": [
15073 - "tensor_descriptor",
15074 - "pointer",
15075 - "pointer",
15076 - "pointer"
15077 - ],
15078 - "pid_type": "flat"
15079 - },
15080 - "intermediate_2880_numtokens_8": {
15081 - "block_sizes": [
15082 - 8,
15083 - 32
15084 - ],
15085 - "loop_orders": [
15086 - [
15087 - 0,
15088 - 1
15089 - ]
15090 - ],
15091 - "flatten_loops": [
15092 - true
15093 - ],
15094 - "l2_groupings": [
15095 - 1
15096 - ],
15097 - "range_unroll_factors": [
15098 - 0
15099 - ],
15100 - "range_warp_specializes": [],
15101 - "range_num_stages": [
15102 - 0
15103 - ],
15104 - "range_multi_buffers": [
15105 - null
15106 - ],
15107 - "range_flattens": [
15108 - null
15109 - ],
15110 - "load_eviction_policies": [
15111 - "",
15112 - "",
15113 - ""
15114 - ],
15115 - "num_warps": 4,
15116 - "num_stages": 1,
15117 - "indexing": [
15118 - "pointer",
15119 - "pointer",
15120 - "pointer",
15121 - "pointer"
15122 - ],
15123 - "pid_type": "flat"
15124 - },
15125 - "intermediate_4096_numtokens_8": {
15126 - "block_sizes": [
15127 - 2,
15128 - 32
15129 - ],
15130 - "loop_orders": [
15131 - [
15132 - 0,
15133 - 1
15134 - ]
15135 - ],
15136 - "flatten_loops": [
15137 - true
15138 - ],
15139 - "l2_groupings": [
15140 - 2
15141 - ],
15142 - "range_unroll_factors": [
15143 - 0
15144 - ],
15145 - "range_warp_specializes": [],
15146 - "range_num_stages": [
15147 - 0
15148 - ],
15149 - "range_multi_buffers": [
15150 - null
15151 - ],
15152 - "range_flattens": [
15153 - null
15154 - ],
15155 - "load_eviction_policies": [
15156 - "",
15157 - "",
15158 - ""
15159 - ],
15160 - "num_warps": 32,
15161 - "num_stages": 1,
15162 - "indexing": [
15163 - "pointer",
15164 - "pointer",
15165 - "pointer",
15166 - "pointer"
15167 - ],
15168 - "pid_type": "flat"
15169 - },
15170 - "intermediate_8192_numtokens_8": {
15171 - "block_sizes": [
15172 - 4,
15173 - 64
15174 - ],
15175 - "loop_orders": [
15176 - [
15177 - 0,
15178 - 1
15179 - ]
15180 - ],
15181 - "flatten_loops": [
15182 - false
15183 - ],
15184 - "l2_groupings": [
15185 - 1
15186 - ],
15187 - "range_unroll_factors": [
15188 - 0
15189 - ],
15190 - "range_warp_specializes": [],
15191 - "range_num_stages": [
15192 - 0
15193 - ],
15194 - "range_multi_buffers": [
15195 - null
15196 - ],
15197 - "range_flattens": [
15198 - null
15199 - ],
15200 - "load_eviction_policies": [
15201 - "",
15202 - "",
15203 - ""
15204 - ],
15205 - "num_warps": 1,
15206 - "num_stages": 1,
15207 - "indexing": [
15208 - "pointer",
15209 - "pointer",
15210 - "pointer",
15211 - "pointer"
15212 - ],
15213 - "pid_type": "flat"
15214 - },
15215 - "intermediate_11008_numtokens_8": {
15216 - "block_sizes": [
15217 - 8,
15218 - 128
15219 - ],
15220 - "loop_orders": [
15221 - [
15222 - 0,
15223 - 1
15224 - ]
15225 - ],
15226 - "flatten_loops": [
15227 - false
15228 - ],
15229 - "l2_groupings": [
15230 - 1
15231 - ],
15232 - "range_unroll_factors": [
15233 - 0
15234 - ],
15235 - "range_warp_specializes": [],
15236 - "range_num_stages": [
15237 - 0
15238 - ],
15239 - "range_multi_buffers": [
15240 - null
15241 - ],
15242 - "range_flattens": [
15243 - null
15244 - ],
15245 - "load_eviction_policies": [
15246 - "last",
15247 - "",
15248 - ""
15249 - ],
15250 - "num_warps": 4,
15251 - "num_stages": 2,
15252 - "indexing": [
15253 - "pointer",
15254 - "pointer",
15255 - "pointer",
15256 - "pointer"
15257 - ],
15258 - "pid_type": "flat"
15259 - },
15260 - "intermediate_14336_numtokens_8": {
15261 - "block_sizes": [
15262 - 8,
15263 - 32
15264 - ],
15265 - "loop_orders": [
15266 - [
15267 - 0,
15268 - 1
15269 - ]
15270 - ],
15271 - "flatten_loops": [
15272 - false
15273 - ],
15274 - "l2_groupings": [
15275 - 1
15276 - ],
15277 - "range_unroll_factors": [
15278 - 0
15279 - ],
15280 - "range_warp_specializes": [],
15281 - "range_num_stages": [
15282 - 0
15283 - ],
15284 - "range_multi_buffers": [
15285 - null
15286 - ],
15287 - "range_flattens": [
15288 - null
15289 - ],
15290 - "load_eviction_policies": [
15291 - "",
15292 - "first",
15293 - ""
15294 - ],
15295 - "num_warps": 4,
15296 - "num_stages": 1,
15297 - "indexing": [
15298 - "pointer",
15299 - "pointer",
15300 - "pointer",
15301 - "pointer"
15302 - ],
15303 - "pid_type": "flat"
15304 - },
15305 - "intermediate_2048_numtokens_16": {
15306 - "block_sizes": [
15307 - 16,
15308 - 64
15309 - ],
15310 - "loop_orders": [
15311 - [
15312 - 0,
15313 - 1
15314 - ]
15315 - ],
15316 - "flatten_loops": [
15317 - false
15318 - ],
15319 - "l2_groupings": [
15320 - 2
15321 - ],
15322 - "range_unroll_factors": [
15323 - 0
15324 - ],
15325 - "range_warp_specializes": [],
15326 - "range_num_stages": [
15327 - 0
15328 - ],
15329 - "range_multi_buffers": [
15330 - null
15331 - ],
15332 - "range_flattens": [
15333 - null
15334 - ],
15335 - "load_eviction_policies": [
15336 - "",
15337 - "",
15338 - ""
15339 - ],
15340 - "num_warps": 4,
15341 - "num_stages": 2,
15342 - "indexing": [
15343 - "pointer",
15344 - "pointer",
15345 - "pointer",
15346 - "pointer"
15347 - ],
15348 - "pid_type": "flat"
15349 - },
15350 - "intermediate_2880_numtokens_16": {
15351 - "block_sizes": [
15352 - 16,
15353 - 32
15354 - ],
15355 - "loop_orders": [
15356 - [
15357 - 0,
15358 - 1
15359 - ]
15360 - ],
15361 - "flatten_loops": [
15362 - true
15363 - ],
15364 - "l2_groupings": [
15365 - 1
15366 - ],
15367 - "range_unroll_factors": [
15368 - 0
15369 - ],
15370 - "range_warp_specializes": [],
15371 - "range_num_stages": [
15372 - 0
15373 - ],
15374 - "range_multi_buffers": [
15375 - null
15376 - ],
15377 - "range_flattens": [
15378 - null
15379 - ],
15380 - "load_eviction_policies": [
15381 - "",
15382 - "",
15383 - ""
15384 - ],
15385 - "num_warps": 8,
15386 - "num_stages": 1,
15387 - "indexing": [
15388 - "pointer",
15389 - "pointer",
15390 - "pointer",
15391 - "pointer"
15392 - ],
15393 - "pid_type": "flat"
15394 - },
15395 - "intermediate_4096_numtokens_16": {
15396 - "block_sizes": [
15397 - 16,
15398 - 32
15399 - ],
15400 - "loop_orders": [
15401 - [
15402 - 0,
15403 - 1
15404 - ]
15405 - ],
15406 - "flatten_loops": [
15407 - true
15408 - ],
15409 - "l2_groupings": [
15410 - 1
15411 - ],
15412 - "range_unroll_factors": [
15413 - 0
15414 - ],
15415 - "range_warp_specializes": [],
15416 - "range_num_stages": [
15417 - 0
15418 - ],
15419 - "range_multi_buffers": [
15420 - null
15421 - ],
15422 - "range_flattens": [
15423 - null
15424 - ],
15425 - "load_eviction_policies": [
15426 - "",
15427 - "",
15428 - ""
15429 - ],
15430 - "num_warps": 8,
15431 - "num_stages": 2,
15432 - "indexing": [
15433 - "pointer",
15434 - "pointer",
15435 - "pointer",
15436 - "pointer"
15437 - ],
15438 - "pid_type": "flat"
15439 - },
15440 - "intermediate_8192_numtokens_16": {
15441 - "block_sizes": [
15442 - 4,
15443 - 32
15444 - ],
15445 - "loop_orders": [
15446 - [
15447 - 0,
15448 - 1
15449 - ]
15450 - ],
15451 - "flatten_loops": [
15452 - false
15453 - ],
15454 - "l2_groupings": [
15455 - 1
15456 - ],
15457 - "range_unroll_factors": [
15458 - 0
15459 - ],
15460 - "range_warp_specializes": [],
15461 - "range_num_stages": [
15462 - 0
15463 - ],
15464 - "range_multi_buffers": [
15465 - null
15466 - ],
15467 - "range_flattens": [
15468 - null
15469 - ],
15470 - "load_eviction_policies": [
15471 - "",
15472 - "",
15473 - ""
15474 - ],
15475 - "num_warps": 1,
15476 - "num_stages": 1,
15477 - "indexing": [
15478 - "pointer",
15479 - "tensor_descriptor",
15480 - "pointer",
15481 - "pointer"
15482 - ],
15483 - "pid_type": "flat"
15484 - },
15485 - "intermediate_11008_numtokens_16": {
15486 - "block_sizes": [
15487 - 8,
15488 - 32
15489 - ],
15490 - "loop_orders": [
15491 - [
15492 - 0,
15493 - 1
15494 - ]
15495 - ],
15496 - "flatten_loops": [
15497 - true
15498 - ],
15499 - "l2_groupings": [
15500 - 1
15501 - ],
15502 - "range_unroll_factors": [
15503 - 0
15504 - ],
15505 - "range_warp_specializes": [],
15506 - "range_num_stages": [
15507 - 0
15508 - ],
15509 - "range_multi_buffers": [
15510 - null
15511 - ],
15512 - "range_flattens": [
15513 - null
15514 - ],
15515 - "load_eviction_policies": [
15516 - "",
15517 - "",
15518 - "first"
15519 - ],
15520 - "num_warps": 16,
15521 - "num_stages": 1,
15522 - "indexing": [
15523 - "pointer",
15524 - "pointer",
15525 - "pointer",
15526 - "pointer"
15527 - ],
15528 - "pid_type": "flat"
15529 - },
15530 - "intermediate_14336_numtokens_16": {
15531 - "block_sizes": [
15532 - 16,
15533 - 32
15534 - ],
15535 - "loop_orders": [
15536 - [
15537 - 0,
15538 - 1
15539 - ]
15540 - ],
15541 - "flatten_loops": [
15542 - true
15543 - ],
15544 - "l2_groupings": [
15545 - 1
15546 - ],
15547 - "range_unroll_factors": [
15548 - 0
15549 - ],
15550 - "range_warp_specializes": [],
15551 - "range_num_stages": [
15552 - 0
15553 - ],
15554 - "range_multi_buffers": [
15555 - null
15556 - ],
15557 - "range_flattens": [
15558 - null
15559 - ],
15560 - "load_eviction_policies": [
15561 - "",
15562 - "",
15563 - ""
15564 - ],
15565 - "num_warps": 1,
15566 - "num_stages": 1,
15567 - "indexing": [
15568 - "tensor_descriptor",
15569 - "pointer",
15570 - "pointer",
15571 - "pointer"
15572 - ],
15573 - "pid_type": "flat"
15574 - },
15575 - "intermediate_2048_numtokens_24": {
15576 - "block_sizes": [
15577 - 16,
15578 - 8
15579 - ],
15580 - "loop_orders": [
15581 - [
15582 - 0,
15583 - 1
15584 - ]
15585 - ],
15586 - "flatten_loops": [
15587 - true
15588 - ],
15589 - "l2_groupings": [
15590 - 1
15591 - ],
15592 - "range_unroll_factors": [
15593 - 0
15594 - ],
15595 - "range_warp_specializes": [],
15596 - "range_num_stages": [
15597 - 0
15598 - ],
15599 - "range_multi_buffers": [
15600 - null
15601 - ],
15602 - "range_flattens": [
15603 - null
15604 - ],
15605 - "load_eviction_policies": [
15606 - "",
15607 - "last",
15608 - "last"
15609 - ],
15610 - "num_warps": 1,
15611 - "num_stages": 1,
15612 - "indexing": [
15613 - "pointer",
15614 - "pointer",
15615 - "pointer",
15616 - "pointer"
15617 - ],
15618 - "pid_type": "flat"
15619 - },
15620 - "intermediate_2880_numtokens_24": {
15621 - "block_sizes": [
15622 - 32,
15623 - 64
15624 - ],
15625 - "loop_orders": [
15626 - [
15627 - 1,
15628 - 0
15629 - ]
15630 - ],
15631 - "flatten_loops": [
15632 - true
15633 - ],
15634 - "l2_groupings": [
15635 - 1
15636 - ],
15637 - "range_unroll_factors": [
15638 - 0
15639 - ],
15640 - "range_warp_specializes": [],
15641 - "range_num_stages": [
15642 - 0
15643 - ],
15644 - "range_multi_buffers": [
15645 - null
15646 - ],
15647 - "range_flattens": [
15648 - null
15649 - ],
15650 - "load_eviction_policies": [
15651 - "",
15652 - "",
15653 - ""
15654 - ],
15655 - "num_warps": 4,
15656 - "num_stages": 1,
15657 - "indexing": [
15658 - "tensor_descriptor",
15659 - "pointer",
15660 - "pointer",
15661 - "tensor_descriptor"
15662 - ],
15663 - "pid_type": "flat"
15664 - },
15665 - "intermediate_4096_numtokens_24": {
15666 - "block_sizes": [
15667 - 32,
15668 - 32
15669 - ],
15670 - "loop_orders": [
15671 - [
15672 - 0,
15673 - 1
15674 - ]
15675 - ],
15676 - "flatten_loops": [
15677 - true
15678 - ],
15679 - "l2_groupings": [
15680 - 2
15681 - ],
15682 - "range_unroll_factors": [
15683 - 0
15684 - ],
15685 - "range_warp_specializes": [],
15686 - "range_num_stages": [
15687 - 0
15688 - ],
15689 - "range_multi_buffers": [
15690 - null
15691 - ],
15692 - "range_flattens": [
15693 - null
15694 - ],
15695 - "load_eviction_policies": [
15696 - "",
15697 - "first",
15698 - ""
15699 - ],
15700 - "num_warps": 1,
15701 - "num_stages": 1,
15702 - "indexing": [
15703 - "pointer",
15704 - "pointer",
15705 - "pointer",
15706 - "tensor_descriptor"
15707 - ],
15708 - "pid_type": "flat"
15709 - },
15710 - "intermediate_8192_numtokens_24": {
15711 - "block_sizes": [
15712 - 16,
15713 - 32
15714 - ],
15715 - "loop_orders": [
15716 - [
15717 - 0,
15718 - 1
15719 - ]
15720 - ],
15721 - "flatten_loops": [
15722 - false
15723 - ],
15724 - "l2_groupings": [
15725 - 1
15726 - ],
15727 - "range_unroll_factors": [
15728 - 0
15729 - ],
15730 - "range_warp_specializes": [],
15731 - "range_num_stages": [
15732 - 0
15733 - ],
15734 - "range_multi_buffers": [
15735 - null
15736 - ],
15737 - "range_flattens": [
15738 - null
15739 - ],
15740 - "load_eviction_policies": [
15741 - "",
15742 - "",
15743 - ""
15744 - ],
15745 - "num_warps": 2,
15746 - "num_stages": 1,
15747 - "indexing": [
15748 - "pointer",
15749 - "pointer",
15750 - "pointer",
15751 - "pointer"
15752 - ],
15753 - "pid_type": "flat"
15754 - },
15755 - "intermediate_11008_numtokens_24": {
15756 - "block_sizes": [
15757 - 32,
15758 - 8
15759 - ],
15760 - "loop_orders": [
15761 - [
15762 - 0,
15763 - 1
15764 - ]
15765 - ],
15766 - "flatten_loops": [
15767 - false
15768 - ],
15769 - "l2_groupings": [
15770 - 1
15771 - ],
15772 - "range_unroll_factors": [
15773 - 0
15774 - ],
15775 - "range_warp_specializes": [],
15776 - "range_num_stages": [
15777 - 0
15778 - ],
15779 - "range_multi_buffers": [
15780 - null
15781 - ],
15782 - "range_flattens": [
15783 - null
15784 - ],
15785 - "load_eviction_policies": [
15786 - "",
15787 - "",
15788 - ""
15789 - ],
15790 - "num_warps": 1,
15791 - "num_stages": 1,
15792 - "indexing": [
15793 - "pointer",
15794 - "pointer",
15795 - "pointer",
15796 - "pointer"
15797 - ],
15798 - "pid_type": "flat"
15799 - },
15800 - "intermediate_14336_numtokens_24": {
15801 - "block_sizes": [
15802 - 8,
15803 - 32
15804 - ],
15805 - "loop_orders": [
15806 - [
15807 - 1,
15808 - 0
15809 - ]
15810 - ],
15811 - "flatten_loops": [
15812 - true
15813 - ],
15814 - "l2_groupings": [
15815 - 1
15816 - ],
15817 - "range_unroll_factors": [
15818 - 0
15819 - ],
15820 - "range_warp_specializes": [],
15821 - "range_num_stages": [
15822 - 0
15823 - ],
15824 - "range_multi_buffers": [
15825 - null
15826 - ],
15827 - "range_flattens": [
15828 - null
15829 - ],
15830 - "load_eviction_policies": [
15831 - "",
15832 - "",
15833 - "last"
15834 - ],
15835 - "num_warps": 2,
15836 - "num_stages": 1,
15837 - "indexing": [
15838 - "pointer",
15839 - "pointer",
15840 - "pointer",
15841 - "pointer"
15842 - ],
15843 - "pid_type": "flat"
15844 - },
15845 - "intermediate_2048_numtokens_32": {
15846 - "block_sizes": [
15847 - 32,
15848 - 64
15849 - ],
15850 - "loop_orders": [
15851 - [
15852 - 0,
15853 - 1
15854 - ]
15855 - ],
15856 - "flatten_loops": [
15857 - true
15858 - ],
15859 - "l2_groupings": [
15860 - 1
15861 - ],
15862 - "range_unroll_factors": [
15863 - 0
15864 - ],
15865 - "range_warp_specializes": [],
15866 - "range_num_stages": [
15867 - 0
15868 - ],
15869 - "range_multi_buffers": [
15870 - null
15871 - ],
15872 - "range_flattens": [
15873 - null
15874 - ],
15875 - "load_eviction_policies": [
15876 - "",
15877 - "",
15878 - ""
15879 - ],
15880 - "num_warps": 1,
15881 - "num_stages": 1,
15882 - "indexing": [
15883 - "pointer",
15884 - "pointer",
15885 - "pointer",
15886 - "pointer"
15887 - ],
15888 - "pid_type": "flat"
15889 - },
15890 - "intermediate_2880_numtokens_32": {
15891 - "block_sizes": [
15892 - 32,
15893 - 64
15894 - ],
15895 - "loop_orders": [
15896 - [
15897 - 0,
15898 - 1
15899 - ]
15900 - ],
15901 - "flatten_loops": [
15902 - true
15903 - ],
15904 - "l2_groupings": [
15905 - 2
15906 - ],
15907 - "range_unroll_factors": [
15908 - 0
15909 - ],
15910 - "range_warp_specializes": [],
15911 - "range_num_stages": [
15912 - 0
15913 - ],
15914 - "range_multi_buffers": [
15915 - null
15916 - ],
15917 - "range_flattens": [
15918 - null
15919 - ],
15920 - "load_eviction_policies": [
15921 - "",
15922 - "",
15923 - ""
15924 - ],
15925 - "num_warps": 1,
15926 - "num_stages": 1,
15927 - "indexing": [
15928 - "pointer",
15929 - "pointer",
15930 - "pointer",
15931 - "pointer"
15932 - ],
15933 - "pid_type": "flat"
15934 - },
15935 - "intermediate_4096_numtokens_32": {
15936 - "block_sizes": [
15937 - 32,
15938 - 16
15939 - ],
15940 - "loop_orders": [
15941 - [
15942 - 0,
15943 - 1
15944 - ]
15945 - ],
15946 - "flatten_loops": [
15947 - true
15948 - ],
15949 - "l2_groupings": [
15950 - 1
15951 - ],
15952 - "range_unroll_factors": [
15953 - 0
15954 - ],
15955 - "range_warp_specializes": [],
15956 - "range_num_stages": [
15957 - 0
15958 - ],
15959 - "range_multi_buffers": [
15960 - null
15961 - ],
15962 - "range_flattens": [
15963 - null
15964 - ],
15965 - "load_eviction_policies": [
15966 - "",
15967 - "",
15968 - ""
15969 - ],
15970 - "num_warps": 1,
15971 - "num_stages": 2,
15972 - "indexing": [
15973 - "pointer",
15974 - "pointer",
15975 - "pointer",
15976 - "pointer"
15977 - ],
15978 - "pid_type": "flat"
15979 - },
15980 - "intermediate_8192_numtokens_32": {
15981 - "block_sizes": [
15982 - 32,
15983 - 128
15984 - ],
15985 - "loop_orders": [
15986 - [
15987 - 0,
15988 - 1
15989 - ]
15990 - ],
15991 - "flatten_loops": [
15992 - true
15993 - ],
15994 - "l2_groupings": [
15995 - 1
15996 - ],
15997 - "range_unroll_factors": [
15998 - 0
15999 - ],
16000 - "range_warp_specializes": [],
16001 - "range_num_stages": [
16002 - 0
16003 - ],
16004 - "range_multi_buffers": [
16005 - null
16006 - ],
16007 - "range_flattens": [
16008 - null
16009 - ],
16010 - "load_eviction_policies": [
16011 - "",
16012 - "",
16013 - "last"
16014 - ],
16015 - "num_warps": 2,
16016 - "num_stages": 2,
16017 - "indexing": [
16018 - "pointer",
16019 - "pointer",
16020 - "pointer",
16021 - "pointer"
16022 - ],
16023 - "pid_type": "flat"
16024 - },
16025 - "intermediate_11008_numtokens_32": {
16026 - "block_sizes": [
16027 - 16,
16028 - 8
16029 - ],
16030 - "loop_orders": [
16031 - [
16032 - 1,
16033 - 0
16034 - ]
16035 - ],
16036 - "flatten_loops": [
16037 - true
16038 - ],
16039 - "l2_groupings": [
16040 - 1
16041 - ],
16042 - "range_unroll_factors": [
16043 - 0
16044 - ],
16045 - "range_warp_specializes": [],
16046 - "range_num_stages": [
16047 - 0
16048 - ],
16049 - "range_multi_buffers": [
16050 - null
16051 - ],
16052 - "range_flattens": [
16053 - null
16054 - ],
16055 - "load_eviction_policies": [
16056 - "",
16057 - "",
16058 - ""
16059 - ],
16060 - "num_warps": 16,
16061 - "num_stages": 1,
16062 - "indexing": [
16063 - "pointer",
16064 - "pointer",
16065 - "pointer",
16066 - "tensor_descriptor"
16067 - ],
16068 - "pid_type": "flat"
16069 - },
16070 - "intermediate_14336_numtokens_32": {
16071 - "block_sizes": [
16072 - 32,
16073 - 8
16074 - ],
16075 - "loop_orders": [
16076 - [
16077 - 0,
16078 - 1
16079 - ]
16080 - ],
16081 - "flatten_loops": [
16082 - true
16083 - ],
16084 - "l2_groupings": [
16085 - 1
16086 - ],
16087 - "range_unroll_factors": [
16088 - 0
16089 - ],
16090 - "range_warp_specializes": [],
16091 - "range_num_stages": [
16092 - 0
16093 - ],
16094 - "range_multi_buffers": [
16095 - null
16096 - ],
16097 - "range_flattens": [
16098 - null
16099 - ],
16100 - "load_eviction_policies": [
16101 - "",
16102 - "",
16103 - ""
16104 - ],
16105 - "num_warps": 8,
16106 - "num_stages": 2,
16107 - "indexing": [
16108 - "pointer",
16109 - "pointer",
16110 - "pointer",
16111 - "pointer"
16112 - ],
16113 - "pid_type": "flat"
16114 - },
16115 - "intermediate_2048_numtokens_40": {
16116 - "block_sizes": [
16117 - 64,
16118 - 32
16119 - ],
16120 - "loop_orders": [
16121 - [
16122 - 0,
16123 - 1
16124 - ]
16125 - ],
16126 - "flatten_loops": [
16127 - true
16128 - ],
16129 - "l2_groupings": [
16130 - 2
16131 - ],
16132 - "range_unroll_factors": [
16133 - 0
16134 - ],
16135 - "range_warp_specializes": [],
16136 - "range_num_stages": [
16137 - 0
16138 - ],
16139 - "range_multi_buffers": [
16140 - null
16141 - ],
16142 - "range_flattens": [
16143 - null
16144 - ],
16145 - "load_eviction_policies": [
16146 - "",
16147 - "",
16148 - ""
16149 - ],
16150 - "num_warps": 2,
16151 - "num_stages": 1,
16152 - "indexing": [
16153 - "pointer",
16154 - "pointer",
16155 - "pointer",
16156 - "pointer"
16157 - ],
16158 - "pid_type": "flat"
16159 - },
16160 - "intermediate_2880_numtokens_40": {
16161 - "block_sizes": [
16162 - 64,
16163 - 32
16164 - ],
16165 - "loop_orders": [
16166 - [
16167 - 0,
16168 - 1
16169 - ]
16170 - ],
16171 - "flatten_loops": [
16172 - true
16173 - ],
16174 - "l2_groupings": [
16175 - 1
16176 - ],
16177 - "range_unroll_factors": [
16178 - 0
16179 - ],
16180 - "range_warp_specializes": [],
16181 - "range_num_stages": [
16182 - 0
16183 - ],
16184 - "range_multi_buffers": [
16185 - null
16186 - ],
16187 - "range_flattens": [
16188 - null
16189 - ],
16190 - "load_eviction_policies": [
16191 - "",
16192 - "",
16193 - "last"
16194 - ],
16195 - "num_warps": 8,
16196 - "num_stages": 2,
16197 - "indexing": [
16198 - "pointer",
16199 - "pointer",
16200 - "pointer",
16201 - "pointer"
16202 - ],
16203 - "pid_type": "flat"
16204 - },
16205 - "intermediate_4096_numtokens_40": {
16206 - "block_sizes": [
16207 - 32,
16208 - 32
16209 - ],
16210 - "loop_orders": [
16211 - [
16212 - 1,
16213 - 0
16214 - ]
16215 - ],
16216 - "flatten_loops": [
16217 - true
16218 - ],
16219 - "l2_groupings": [
16220 - 1
16221 - ],
16222 - "range_unroll_factors": [
16223 - 0
16224 - ],
16225 - "range_warp_specializes": [],
16226 - "range_num_stages": [
16227 - 0
16228 - ],
16229 - "range_multi_buffers": [
16230 - null
16231 - ],
16232 - "range_flattens": [
16233 - null
16234 - ],
16235 - "load_eviction_policies": [
16236 - "",
16237 - "",
16238 - ""
16239 - ],
16240 - "num_warps": 8,
16241 - "num_stages": 1,
16242 - "indexing": [
16243 - "tensor_descriptor",
16244 - "pointer",
16245 - "pointer",
16246 - "pointer"
16247 - ],
16248 - "pid_type": "flat"
16249 - },
16250 - "intermediate_8192_numtokens_40": {
16251 - "block_sizes": [
16252 - 64,
16253 - 128
16254 - ],
16255 - "loop_orders": [
16256 - [
16257 - 0,
16258 - 1
16259 - ]
16260 - ],
16261 - "flatten_loops": [
16262 - true
16263 - ],
16264 - "l2_groupings": [
16265 - 2
16266 - ],
16267 - "range_unroll_factors": [
16268 - 0
16269 - ],
16270 - "range_warp_specializes": [],
16271 - "range_num_stages": [
16272 - 0
16273 - ],
16274 - "range_multi_buffers": [
16275 - null
16276 - ],
16277 - "range_flattens": [
16278 - null
16279 - ],
16280 - "load_eviction_policies": [
16281 - "",
16282 - "",
16283 - ""
16284 - ],
16285 - "num_warps": 4,
16286 - "num_stages": 1,
16287 - "indexing": [
16288 - "pointer",
16289 - "pointer",
16290 - "pointer",
16291 - "pointer"
16292 - ],
16293 - "pid_type": "flat"
16294 - },
16295 - "intermediate_11008_numtokens_40": {
16296 - "block_sizes": [
16297 - 64,
16298 - 32
16299 - ],
16300 - "loop_orders": [
16301 - [
16302 - 0,
16303 - 1
16304 - ]
16305 - ],
16306 - "flatten_loops": [
16307 - true
16308 - ],
16309 - "l2_groupings": [
16310 - 1
16311 - ],
16312 - "range_unroll_factors": [
16313 - 0
16314 - ],
16315 - "range_warp_specializes": [],
16316 - "range_num_stages": [
16317 - 0
16318 - ],
16319 - "range_multi_buffers": [
16320 - null
16321 - ],
16322 - "range_flattens": [
16323 - null
16324 - ],
16325 - "load_eviction_policies": [
16326 - "",
16327 - "",
16328 - ""
16329 - ],
16330 - "num_warps": 4,
16331 - "num_stages": 1,
16332 - "indexing": [
16333 - "pointer",
16334 - "pointer",
16335 - "pointer",
16336 - "pointer"
16337 - ],
16338 - "pid_type": "flat"
16339 - },
16340 - "intermediate_14336_numtokens_40": {
16341 - "block_sizes": [
16342 - 32,
16343 - 32
16344 - ],
16345 - "loop_orders": [
16346 - [
16347 - 0,
16348 - 1
16349 - ]
16350 - ],
16351 - "flatten_loops": [
16352 - true
16353 - ],
16354 - "l2_groupings": [
16355 - 1
16356 - ],
16357 - "range_unroll_factors": [
16358 - 0
16359 - ],
16360 - "range_warp_specializes": [],
16361 - "range_num_stages": [
16362 - 0
16363 - ],
16364 - "range_multi_buffers": [
16365 - null
16366 - ],
16367 - "range_flattens": [
16368 - null
16369 - ],
16370 - "load_eviction_policies": [
16371 - "",
16372 - "",
16373 - ""
16374 - ],
16375 - "num_warps": 2,
16376 - "num_stages": 1,
16377 - "indexing": [
16378 - "pointer",
16379 - "pointer",
16380 - "pointer",
16381 - "pointer"
16382 - ],
16383 - "pid_type": "flat"
16384 - },
16385 - "intermediate_2048_numtokens_48": {
16386 - "block_sizes": [
16387 - 32,
16388 - 8
16389 - ],
16390 - "loop_orders": [
16391 - [
16392 - 0,
16393 - 1
16394 - ]
16395 - ],
16396 - "flatten_loops": [
16397 - true
16398 - ],
16399 - "l2_groupings": [
16400 - 2
16401 - ],
16402 - "range_unroll_factors": [
16403 - 0
16404 - ],
16405 - "range_warp_specializes": [],
16406 - "range_num_stages": [
16407 - 0
16408 - ],
16409 - "range_multi_buffers": [
16410 - null
16411 - ],
16412 - "range_flattens": [
16413 - null
16414 - ],
16415 - "load_eviction_policies": [
16416 - "",
16417 - "first",
16418 - ""
16419 - ],
16420 - "num_warps": 16,
16421 - "num_stages": 2,
16422 - "indexing": [
16423 - "pointer",
16424 - "pointer",
16425 - "tensor_descriptor",
16426 - "pointer"
16427 - ],
16428 - "pid_type": "flat"
16429 - },
16430 - "intermediate_2880_numtokens_48": {
16431 - "block_sizes": [
16432 - 8,
16433 - 8
16434 - ],
16435 - "loop_orders": [
16436 - [
16437 - 0,
16438 - 1
16439 - ]
16440 - ],
16441 - "flatten_loops": [
16442 - true
16443 - ],
16444 - "l2_groupings": [
16445 - 1
16446 - ],
16447 - "range_unroll_factors": [
16448 - 0
16449 - ],
16450 - "range_warp_specializes": [],
16451 - "range_num_stages": [
16452 - 0
16453 - ],
16454 - "range_multi_buffers": [
16455 - null
16456 - ],
16457 - "range_flattens": [
16458 - null
16459 - ],
16460 - "load_eviction_policies": [
16461 - "",
16462 - "",
16463 - ""
16464 - ],
16465 - "num_warps": 1,
16466 - "num_stages": 1,
16467 - "indexing": [
16468 - "pointer",
16469 - "pointer",
16470 - "tensor_descriptor",
16471 - "pointer"
16472 - ],
16473 - "pid_type": "flat"
16474 - },
16475 - "intermediate_4096_numtokens_48": {
16476 - "block_sizes": [
16477 - 32,
16478 - 64
16479 - ],
16480 - "loop_orders": [
16481 - [
16482 - 0,
16483 - 1
16484 - ]
16485 - ],
16486 - "flatten_loops": [
16487 - true
16488 - ],
16489 - "l2_groupings": [
16490 - 1
16491 - ],
16492 - "range_unroll_factors": [
16493 - 0
16494 - ],
16495 - "range_warp_specializes": [],
16496 - "range_num_stages": [
16497 - 0
16498 - ],
16499 - "range_multi_buffers": [
16500 - null
16501 - ],
16502 - "range_flattens": [
16503 - null
16504 - ],
16505 - "load_eviction_policies": [
16506 - "",
16507 - "",
16508 - ""
16509 - ],
16510 - "num_warps": 1,
16511 - "num_stages": 1,
16512 - "indexing": [
16513 - "pointer",
16514 - "pointer",
16515 - "pointer",
16516 - "tensor_descriptor"
16517 - ],
16518 - "pid_type": "flat"
16519 - },
16520 - "intermediate_8192_numtokens_48": {
16521 - "block_sizes": [
16522 - 64,
16523 - 32
16524 - ],
16525 - "loop_orders": [
16526 - [
16527 - 0,
16528 - 1
16529 - ]
16530 - ],
16531 - "flatten_loops": [
16532 - true
16533 - ],
16534 - "l2_groupings": [
16535 - 1
16536 - ],
16537 - "range_unroll_factors": [
16538 - 0
16539 - ],
16540 - "range_warp_specializes": [],
16541 - "range_num_stages": [
16542 - 0
16543 - ],
16544 - "range_multi_buffers": [
16545 - null
16546 - ],
16547 - "range_flattens": [
16548 - null
16549 - ],
16550 - "load_eviction_policies": [
16551 - "",
16552 - "",
16553 - ""
16554 - ],
16555 - "num_warps": 4,
16556 - "num_stages": 1,
16557 - "indexing": [
16558 - "pointer",
16559 - "pointer",
16560 - "pointer",
16561 - "pointer"
16562 - ],
16563 - "pid_type": "flat"
16564 - },
16565 - "intermediate_11008_numtokens_48": {
16566 - "block_sizes": [
16567 - 16,
16568 - 256
16569 - ],
16570 - "loop_orders": [
16571 - [
16572 - 0,
16573 - 1
16574 - ]
16575 - ],
16576 - "flatten_loops": [
16577 - true
16578 - ],
16579 - "l2_groupings": [
16580 - 2
16581 - ],
16582 - "range_unroll_factors": [
16583 - 0
16584 - ],
16585 - "range_warp_specializes": [],
16586 - "range_num_stages": [
16587 - 0
16588 - ],
16589 - "range_multi_buffers": [
16590 - null
16591 - ],
16592 - "range_flattens": [
16593 - null
16594 - ],
16595 - "load_eviction_policies": [
16596 - "",
16597 - "",
16598 - ""
16599 - ],
16600 - "num_warps": 32,
16601 - "num_stages": 1,
16602 - "indexing": [
16603 - "pointer",
16604 - "pointer",
16605 - "pointer",
16606 - "tensor_descriptor"
16607 - ],
16608 - "pid_type": "flat"
16609 - },
16610 - "intermediate_14336_numtokens_48": {
16611 - "block_sizes": [
16612 - 64,
16613 - 4
16614 - ],
16615 - "loop_orders": [
16616 - [
16617 - 0,
16618 - 1
16619 - ]
16620 - ],
16621 - "flatten_loops": [
16622 - true
16623 - ],
16624 - "l2_groupings": [
16625 - 1
16626 - ],
16627 - "range_unroll_factors": [
16628 - 0
16629 - ],
16630 - "range_warp_specializes": [],
16631 - "range_num_stages": [
16632 - 0
16633 - ],
16634 - "range_multi_buffers": [
16635 - null
16636 - ],
16637 - "range_flattens": [
16638 - null
16639 - ],
16640 - "load_eviction_policies": [
16641 - "",
16642 - "",
16643 - ""
16644 - ],
16645 - "num_warps": 2,
16646 - "num_stages": 1,
16647 - "indexing": [
16648 - "tensor_descriptor",
16649 - "pointer",
16650 - "pointer",
16651 - "pointer"
16652 - ],
16653 - "pid_type": "flat"
16654 - },
16655 - "intermediate_2048_numtokens_56": {
16656 - "block_sizes": [
16657 - 2,
16658 - 8
16659 - ],
16660 - "loop_orders": [
16661 - [
16662 - 0,
16663 - 1
16664 - ]
16665 - ],
16666 - "flatten_loops": [
16667 - true
16668 - ],
16669 - "l2_groupings": [
16670 - 1
16671 - ],
16672 - "range_unroll_factors": [
16673 - 0
16674 - ],
16675 - "range_warp_specializes": [],
16676 - "range_num_stages": [
16677 - 0
16678 - ],
16679 - "range_multi_buffers": [
16680 - null
16681 - ],
16682 - "range_flattens": [
16683 - null
16684 - ],
16685 - "load_eviction_policies": [
16686 - "",
16687 - "",
16688 - ""
16689 - ],
16690 - "num_warps": 1,
16691 - "num_stages": 1,
16692 - "indexing": [
16693 - "tensor_descriptor",
16694 - "pointer",
16695 - "tensor_descriptor",
16696 - "pointer"
16697 - ],
16698 - "pid_type": "flat"
16699 - },
16700 - "intermediate_2880_numtokens_56": {
16701 - "block_sizes": [
16702 - 8,
16703 - 32
16704 - ],
16705 - "loop_orders": [
16706 - [
16707 - 0,
16708 - 1
16709 - ]
16710 - ],
16711 - "flatten_loops": [
16712 - true
16713 - ],
16714 - "l2_groupings": [
16715 - 1
16716 - ],
16717 - "range_unroll_factors": [
16718 - 0
16719 - ],
16720 - "range_warp_specializes": [],
16721 - "range_num_stages": [
16722 - 0
16723 - ],
16724 - "range_multi_buffers": [
16725 - null
16726 - ],
16727 - "range_flattens": [
16728 - null
16729 - ],
16730 - "load_eviction_policies": [
16731 - "",
16732 - "",
16733 - ""
16734 - ],
16735 - "num_warps": 4,
16736 - "num_stages": 1,
16737 - "indexing": [
16738 - "pointer",
16739 - "pointer",
16740 - "pointer",
16741 - "pointer"
16742 - ],
16743 - "pid_type": "flat"
16744 - },
16745 - "intermediate_4096_numtokens_56": {
16746 - "block_sizes": [
16747 - 32,
16748 - 4
16749 - ],
16750 - "loop_orders": [
16751 - [
16752 - 0,
16753 - 1
16754 - ]
16755 - ],
16756 - "flatten_loops": [
16757 - true
16758 - ],
16759 - "l2_groupings": [
16760 - 1
16761 - ],
16762 - "range_unroll_factors": [
16763 - 0
16764 - ],
16765 - "range_warp_specializes": [],
16766 - "range_num_stages": [
16767 - 0
16768 - ],
16769 - "range_multi_buffers": [
16770 - null
16771 - ],
16772 - "range_flattens": [
16773 - null
16774 - ],
16775 - "load_eviction_policies": [
16776 - "",
16777 - "first",
16778 - ""
16779 - ],
16780 - "num_warps": 4,
16781 - "num_stages": 1,
16782 - "indexing": [
16783 - "pointer",
16784 - "pointer",
16785 - "tensor_descriptor",
16786 - "pointer"
16787 - ],
16788 - "pid_type": "flat"
16789 - },
16790 - "intermediate_8192_numtokens_56": {
16791 - "block_sizes": [
16792 - 32,
16793 - 128
16794 - ],
16795 - "loop_orders": [
16796 - [
16797 - 0,
16798 - 1
16799 - ]
16800 - ],
16801 - "flatten_loops": [
16802 - true
16803 - ],
16804 - "l2_groupings": [
16805 - 1
16806 - ],
16807 - "range_unroll_factors": [
16808 - 0
16809 - ],
16810 - "range_warp_specializes": [],
16811 - "range_num_stages": [
16812 - 0
16813 - ],
16814 - "range_multi_buffers": [
16815 - null
16816 - ],
16817 - "range_flattens": [
16818 - null
16819 - ],
16820 - "load_eviction_policies": [
16821 - "",
16822 - "",
16823 - ""
16824 - ],
16825 - "num_warps": 1,
16826 - "num_stages": 1,
16827 - "indexing": [
16828 - "pointer",
16829 - "pointer",
16830 - "pointer",
16831 - "pointer"
16832 - ],
16833 - "pid_type": "flat"
16834 - },
16835 - "intermediate_11008_numtokens_56": {
16836 - "block_sizes": [
16837 - 32,
16838 - 8
16839 - ],
16840 - "loop_orders": [
16841 - [
16842 - 0,
16843 - 1
16844 - ]
16845 - ],
16846 - "flatten_loops": [
16847 - true
16848 - ],
16849 - "l2_groupings": [
16850 - 1
16851 - ],
16852 - "range_unroll_factors": [
16853 - 0
16854 - ],
16855 - "range_warp_specializes": [],
16856 - "range_num_stages": [
16857 - 0
16858 - ],
16859 - "range_multi_buffers": [
16860 - null
16861 - ],
16862 - "range_flattens": [
16863 - null
16864 - ],
16865 - "load_eviction_policies": [
16866 - "",
16867 - "first",
16868 - "last"
16869 - ],
16870 - "num_warps": 1,
16871 - "num_stages": 1,
16872 - "indexing": [
16873 - "pointer",
16874 - "pointer",
16875 - "pointer",
16876 - "pointer"
16877 - ],
16878 - "pid_type": "flat"
16879 - },
16880 - "intermediate_14336_numtokens_56": {
16881 - "block_sizes": [
16882 - 64,
16883 - 64
16884 - ],
16885 - "loop_orders": [
16886 - [
16887 - 0,
16888 - 1
16889 - ]
16890 - ],
16891 - "flatten_loops": [
16892 - true
16893 - ],
16894 - "l2_groupings": [
16895 - 2
16896 - ],
16897 - "range_unroll_factors": [
16898 - 0
16899 - ],
16900 - "range_warp_specializes": [],
16901 - "range_num_stages": [
16902 - 0
16903 - ],
16904 - "range_multi_buffers": [
16905 - null
16906 - ],
16907 - "range_flattens": [
16908 - null
16909 - ],
16910 - "load_eviction_policies": [
16911 - "",
16912 - "",
16913 - ""
16914 - ],
16915 - "num_warps": 16,
16916 - "num_stages": 1,
16917 - "indexing": [
16918 - "pointer",
16919 - "pointer",
16920 - "pointer",
16921 - "tensor_descriptor"
16922 - ],
16923 - "pid_type": "flat"
16924 - },
16925 - "intermediate_2048_numtokens_64": {
16926 - "block_sizes": [
16927 - 16,
16928 - 128
16929 - ],
16930 - "loop_orders": [
16931 - [
16932 - 1,
16933 - 0
16934 - ]
16935 - ],
16936 - "flatten_loops": [
16937 - true
16938 - ],
16939 - "l2_groupings": [
16940 - 1
16941 - ],
16942 - "range_unroll_factors": [
16943 - 0
16944 - ],
16945 - "range_warp_specializes": [],
16946 - "range_num_stages": [
16947 - 0
16948 - ],
16949 - "range_multi_buffers": [
16950 - null
16951 - ],
16952 - "range_flattens": [
16953 - null
16954 - ],
16955 - "load_eviction_policies": [
16956 - "",
16957 - "",
16958 - ""
16959 - ],
16960 - "num_warps": 8,
16961 - "num_stages": 1,
16962 - "indexing": [
16963 - "tensor_descriptor",
16964 - "pointer",
16965 - "pointer",
16966 - "pointer"
16967 - ],
16968 - "pid_type": "flat"
16969 - },
16970 - "intermediate_2880_numtokens_64": {
16971 - "block_sizes": [
16972 - 4,
16973 - 64
16974 - ],
16975 - "loop_orders": [
16976 - [
16977 - 1,
16978 - 0
16979 - ]
16980 - ],
16981 - "flatten_loops": [
16982 - true
16983 - ],
16984 - "l2_groupings": [
16985 - 1
16986 - ],
16987 - "range_unroll_factors": [
16988 - 0
16989 - ],
16990 - "range_warp_specializes": [],
16991 - "range_num_stages": [
16992 - 0
16993 - ],
16994 - "range_multi_buffers": [
16995 - null
16996 - ],
16997 - "range_flattens": [
16998 - null
16999 - ],
17000 - "load_eviction_policies": [
17001 - "",
17002 - "",
17003 - ""
17004 - ],
17005 - "num_warps": 4,
17006 - "num_stages": 1,
17007 - "indexing": [
17008 - "pointer",
17009 - "pointer",
17010 - "pointer",
17011 - "pointer"
17012 - ],
17013 - "pid_type": "flat"
17014 - },
17015 - "intermediate_4096_numtokens_64": {
17016 - "block_sizes": [
17017 - 2,
17018 - 16
17019 - ],
17020 - "loop_orders": [
17021 - [
17022 - 0,
17023 - 1
17024 - ]
17025 - ],
17026 - "flatten_loops": [
17027 - true
17028 - ],
17029 - "l2_groupings": [
17030 - 1
17031 - ],
17032 - "range_unroll_factors": [
17033 - 0
17034 - ],
17035 - "range_warp_specializes": [],
17036 - "range_num_stages": [
17037 - 0
17038 - ],
17039 - "range_multi_buffers": [
17040 - null
17041 - ],
17042 - "range_flattens": [
17043 - null
17044 - ],
17045 - "load_eviction_policies": [
17046 - "first",
17047 - "",
17048 - ""
17049 - ],
17050 - "num_warps": 1,
17051 - "num_stages": 1,
17052 - "indexing": [
17053 - "pointer",
17054 - "pointer",
17055 - "pointer",
17056 - "pointer"
17057 - ],
17058 - "pid_type": "flat"
17059 - },
17060 - "intermediate_8192_numtokens_64": {
17061 - "block_sizes": [
17062 - 8,
17063 - 32
17064 - ],
17065 - "loop_orders": [
17066 - [
17067 - 0,
17068 - 1
17069 - ]
17070 - ],
17071 - "flatten_loops": [
17072 - true
17073 - ],
17074 - "l2_groupings": [
17075 - 1
17076 - ],
17077 - "range_unroll_factors": [
17078 - 0
17079 - ],
17080 - "range_warp_specializes": [],
17081 - "range_num_stages": [
17082 - 0
17083 - ],
17084 - "range_multi_buffers": [
17085 - null
17086 - ],
17087 - "range_flattens": [
17088 - null
17089 - ],
17090 - "load_eviction_policies": [
17091 - "",
17092 - "",
17093 - ""
17094 - ],
17095 - "num_warps": 16,
17096 - "num_stages": 2,
17097 - "indexing": [
17098 - "pointer",
17099 - "pointer",
17100 - "pointer",
17101 - "pointer"
17102 - ],
17103 - "pid_type": "flat"
17104 - },
17105 - "intermediate_11008_numtokens_64": {
17106 - "block_sizes": [
17107 - 32,
17108 - 32
17109 - ],
17110 - "loop_orders": [
17111 - [
17112 - 0,
17113 - 1
17114 - ]
17115 - ],
17116 - "flatten_loops": [
17117 - false
17118 - ],
17119 - "l2_groupings": [
17120 - 1
17121 - ],
17122 - "range_unroll_factors": [
17123 - 0
17124 - ],
17125 - "range_warp_specializes": [],
17126 - "range_num_stages": [
17127 - 0
17128 - ],
17129 - "range_multi_buffers": [
17130 - null
17131 - ],
17132 - "range_flattens": [
17133 - null
17134 - ],
17135 - "load_eviction_policies": [
17136 - "",
17137 - "",
17138 - ""
17139 - ],
17140 - "num_warps": 2,
17141 - "num_stages": 1,
17142 - "indexing": [
17143 - "pointer",
17144 - "pointer",
17145 - "tensor_descriptor",
17146 - "pointer"
17147 - ],
17148 - "pid_type": "flat"
17149 - },
17150 - "intermediate_14336_numtokens_64": {
17151 - "block_sizes": [
17152 - 32,
17153 - 8
17154 - ],
17155 - "loop_orders": [
17156 - [
17157 - 0,
17158 - 1
17159 - ]
17160 - ],
17161 - "flatten_loops": [
17162 - true
17163 - ],
17164 - "l2_groupings": [
17165 - 1
17166 - ],
17167 - "range_unroll_factors": [
17168 - 0
17169 - ],
17170 - "range_warp_specializes": [],
17171 - "range_num_stages": [
17172 - 0
17173 - ],
17174 - "range_multi_buffers": [
17175 - null
17176 - ],
17177 - "range_flattens": [
17178 - null
17179 - ],
17180 - "load_eviction_policies": [
17181 - "",
17182 - "",
17183 - ""
17184 - ],
17185 - "num_warps": 16,
17186 - "num_stages": 1,
17187 - "indexing": [
17188 - "pointer",
17189 - "pointer",
17190 - "pointer",
17191 - "pointer"
17192 - ],
17193 - "pid_type": "flat"
17194 - },
17195 - "intermediate_2048_numtokens_72": {
17196 - "block_sizes": [
17197 - 4,
17198 - 16
17199 - ],
17200 - "loop_orders": [
17201 - [
17202 - 0,
17203 - 1
17204 - ]
17205 - ],
17206 - "flatten_loops": [
17207 - true
17208 - ],
17209 - "l2_groupings": [
17210 - 1
17211 - ],
17212 - "range_unroll_factors": [
17213 - 0
17214 - ],
17215 - "range_warp_specializes": [],
17216 - "range_num_stages": [
17217 - 0
17218 - ],
17219 - "range_multi_buffers": [
17220 - null
17221 - ],
17222 - "range_flattens": [
17223 - null
17224 - ],
17225 - "load_eviction_policies": [
17226 - "",
17227 - "",
17228 - "last"
17229 - ],
17230 - "num_warps": 1,
17231 - "num_stages": 1,
17232 - "indexing": [
17233 - "pointer",
17234 - "pointer",
17235 - "pointer",
17236 - "pointer"
17237 - ],
17238 - "pid_type": "flat"
17239 - },
17240 - "intermediate_2880_numtokens_72": {
17241 - "block_sizes": [
17242 - 64,
17243 - 32
17244 - ],
17245 - "loop_orders": [
17246 - [
17247 - 0,
17248 - 1
17249 - ]
17250 - ],
17251 - "flatten_loops": [
17252 - true
17253 - ],
17254 - "l2_groupings": [
17255 - 1
17256 - ],
17257 - "range_unroll_factors": [
17258 - 0
17259 - ],
17260 - "range_warp_specializes": [],
17261 - "range_num_stages": [
17262 - 0
17263 - ],
17264 - "range_multi_buffers": [
17265 - null
17266 - ],
17267 - "range_flattens": [
17268 - null
17269 - ],
17270 - "load_eviction_policies": [
17271 - "",
17272 - "",
17273 - "first"
17274 - ],
17275 - "num_warps": 16,
17276 - "num_stages": 1,
17277 - "indexing": [
17278 - "pointer",
17279 - "pointer",
17280 - "pointer",
17281 - "pointer"
17282 - ],
17283 - "pid_type": "flat"
17284 - },
17285 - "intermediate_4096_numtokens_72": {
17286 - "block_sizes": [
17287 - 64,
17288 - 16
17289 - ],
17290 - "loop_orders": [
17291 - [
17292 - 1,
17293 - 0
17294 - ]
17295 - ],
17296 - "flatten_loops": [
17297 - true
17298 - ],
17299 - "l2_groupings": [
17300 - 1
17301 - ],
17302 - "range_unroll_factors": [
17303 - 0
17304 - ],
17305 - "range_warp_specializes": [],
17306 - "range_num_stages": [
17307 - 0
17308 - ],
17309 - "range_multi_buffers": [
17310 - null
17311 - ],
17312 - "range_flattens": [
17313 - null
17314 - ],
17315 - "load_eviction_policies": [
17316 - "last",
17317 - "",
17318 - ""
17319 - ],
17320 - "num_warps": 32,
17321 - "num_stages": 1,
17322 - "indexing": [
17323 - "pointer",
17324 - "pointer",
17325 - "pointer",
17326 - "pointer"
17327 - ],
17328 - "pid_type": "flat"
17329 - },
17330 - "intermediate_8192_numtokens_72": {
17331 - "block_sizes": [
17332 - 32,
17333 - 8
17334 - ],
17335 - "loop_orders": [
17336 - [
17337 - 0,
17338 - 1
17339 - ]
17340 - ],
17341 - "flatten_loops": [
17342 - true
17343 - ],
17344 - "l2_groupings": [
17345 - 1
17346 - ],
17347 - "range_unroll_factors": [
17348 - 0
17349 - ],
17350 - "range_warp_specializes": [],
17351 - "range_num_stages": [
17352 - 0
17353 - ],
17354 - "range_multi_buffers": [
17355 - null
17356 - ],
17357 - "range_flattens": [
17358 - null
17359 - ],
17360 - "load_eviction_policies": [
17361 - "",
17362 - "",
17363 - ""
17364 - ],
17365 - "num_warps": 32,
17366 - "num_stages": 1,
17367 - "indexing": [
17368 - "pointer",
17369 - "pointer",
17370 - "pointer",
17371 - "pointer"
17372 - ],
17373 - "pid_type": "flat"
17374 - },
17375 - "intermediate_11008_numtokens_72": {
17376 - "block_sizes": [
17377 - 32,
17378 - 32
17379 - ],
17380 - "loop_orders": [
17381 - [
17382 - 0,
17383 - 1
17384 - ]
17385 - ],
17386 - "flatten_loops": [
17387 - true
17388 - ],
17389 - "l2_groupings": [
17390 - 1
17391 - ],
17392 - "range_unroll_factors": [
17393 - 0
17394 - ],
17395 - "range_warp_specializes": [],
17396 - "range_num_stages": [
17397 - 0
17398 - ],
17399 - "range_multi_buffers": [
17400 - null
17401 - ],
17402 - "range_flattens": [
17403 - null
17404 - ],
17405 - "load_eviction_policies": [
17406 - "",
17407 - "",
17408 - ""
17409 - ],
17410 - "num_warps": 1,
17411 - "num_stages": 2,
17412 - "indexing": [
17413 - "pointer",
17414 - "pointer",
17415 - "tensor_descriptor",
17416 - "pointer"
17417 - ],
17418 - "pid_type": "flat"
17419 - },
17420 - "intermediate_14336_numtokens_72": {
17421 - "block_sizes": [
17422 - 128,
17423 - 128
17424 - ],
17425 - "loop_orders": [
17426 - [
17427 - 0,
17428 - 1
17429 - ]
17430 - ],
17431 - "flatten_loops": [
17432 - true
17433 - ],
17434 - "l2_groupings": [
17435 - 1
17436 - ],
17437 - "range_unroll_factors": [
17438 - 0
17439 - ],
17440 - "range_warp_specializes": [],
17441 - "range_num_stages": [
17442 - 0
17443 - ],
17444 - "range_multi_buffers": [
17445 - null
17446 - ],
17447 - "range_flattens": [
17448 - null
17449 - ],
17450 - "load_eviction_policies": [
17451 - "",
17452 - "",
17453 - ""
17454 - ],
17455 - "num_warps": 16,
17456 - "num_stages": 1,
17457 - "indexing": [
17458 - "pointer",
17459 - "pointer",
17460 - "pointer",
17461 - "pointer"
17462 - ],
17463 - "pid_type": "flat"
17464 - },
17465 - "intermediate_2048_numtokens_80": {
17466 - "block_sizes": [
17467 - 32,
17468 - 64
17469 - ],
17470 - "loop_orders": [
17471 - [
17472 - 0,
17473 - 1
17474 - ]
17475 - ],
17476 - "flatten_loops": [
17477 - true
17478 - ],
17479 - "l2_groupings": [
17480 - 1
17481 - ],
17482 - "range_unroll_factors": [
17483 - 0
17484 - ],
17485 - "range_warp_specializes": [],
17486 - "range_num_stages": [
17487 - 0
17488 - ],
17489 - "range_multi_buffers": [
17490 - null
17491 - ],
17492 - "range_flattens": [
17493 - null
17494 - ],
17495 - "load_eviction_policies": [
17496 - "",
17497 - "",
17498 - ""
17499 - ],
17500 - "num_warps": 4,
17501 - "num_stages": 1,
17502 - "indexing": [
17503 - "pointer",
17504 - "pointer",
17505 - "pointer",
17506 - "pointer"
17507 - ],
17508 - "pid_type": "flat"
17509 - },
17510 - "intermediate_2880_numtokens_80": {
17511 - "block_sizes": [
17512 - 32,
17513 - 128
17514 - ],
17515 - "loop_orders": [
17516 - [
17517 - 0,
17518 - 1
17519 - ]
17520 - ],
17521 - "flatten_loops": [
17522 - true
17523 - ],
17524 - "l2_groupings": [
17525 - 1
17526 - ],
17527 - "range_unroll_factors": [
17528 - 0
17529 - ],
17530 - "range_warp_specializes": [],
17531 - "range_num_stages": [
17532 - 0
17533 - ],
17534 - "range_multi_buffers": [
17535 - null
17536 - ],
17537 - "range_flattens": [
17538 - null
17539 - ],
17540 - "load_eviction_policies": [
17541 - "",
17542 - "",
17543 - ""
17544 - ],
17545 - "num_warps": 16,
17546 - "num_stages": 1,
17547 - "indexing": [
17548 - "pointer",
17549 - "pointer",
17550 - "pointer",
17551 - "pointer"
17552 - ],
17553 - "pid_type": "flat"
17554 - },
17555 - "intermediate_4096_numtokens_80": {
17556 - "block_sizes": [
17557 - 32,
17558 - 32
17559 - ],
17560 - "loop_orders": [
17561 - [
17562 - 0,
17563 - 1
17564 - ]
17565 - ],
17566 - "flatten_loops": [
17567 - true
17568 - ],
17569 - "l2_groupings": [
17570 - 1
17571 - ],
17572 - "range_unroll_factors": [
17573 - 0
17574 - ],
17575 - "range_warp_specializes": [],
17576 - "range_num_stages": [
17577 - 0
17578 - ],
17579 - "range_multi_buffers": [
17580 - null
17581 - ],
17582 - "range_flattens": [
17583 - null
17584 - ],
17585 - "load_eviction_policies": [
17586 - "",
17587 - "",
17588 - ""
17589 - ],
17590 - "num_warps": 2,
17591 - "num_stages": 1,
17592 - "indexing": [
17593 - "pointer",
17594 - "pointer",
17595 - "pointer",
17596 - "pointer"
17597 - ],
17598 - "pid_type": "flat"
17599 - },
17600 - "intermediate_8192_numtokens_80": {
17601 - "block_sizes": [
17602 - 32,
17603 - 8
17604 - ],
17605 - "loop_orders": [
17606 - [
17607 - 0,
17608 - 1
17609 - ]
17610 - ],
17611 - "flatten_loops": [
17612 - true
17613 - ],
17614 - "l2_groupings": [
17615 - 4
17616 - ],
17617 - "range_unroll_factors": [
17618 - 0
17619 - ],
17620 - "range_warp_specializes": [],
17621 - "range_num_stages": [
17622 - 0
17623 - ],
17624 - "range_multi_buffers": [
17625 - null
17626 - ],
17627 - "range_flattens": [
17628 - null
17629 - ],
17630 - "load_eviction_policies": [
17631 - "",
17632 - "",
17633 - ""
17634 - ],
17635 - "num_warps": 1,
17636 - "num_stages": 1,
17637 - "indexing": [
17638 - "tensor_descriptor",
17639 - "pointer",
17640 - "pointer",
17641 - "pointer"
17642 - ],
17643 - "pid_type": "flat"
17644 - },
17645 - "intermediate_11008_numtokens_80": {
17646 - "block_sizes": [
17647 - 64,
17648 - 8
17649 - ],
17650 - "loop_orders": [
17651 - [
17652 - 0,
17653 - 1
17654 - ]
17655 - ],
17656 - "flatten_loops": [
17657 - true
17658 - ],
17659 - "l2_groupings": [
17660 - 1
17661 - ],
17662 - "range_unroll_factors": [
17663 - 0
17664 - ],
17665 - "range_warp_specializes": [],
17666 - "range_num_stages": [
17667 - 0
17668 - ],
17669 - "range_multi_buffers": [
17670 - null
17671 - ],
17672 - "range_flattens": [
17673 - null
17674 - ],
17675 - "load_eviction_policies": [
17676 - "",
17677 - "",
17678 - ""
17679 - ],
17680 - "num_warps": 8,
17681 - "num_stages": 1,
17682 - "indexing": [
17683 - "pointer",
17684 - "pointer",
17685 - "pointer",
17686 - "pointer"
17687 - ],
17688 - "pid_type": "flat"
17689 - },
17690 - "intermediate_14336_numtokens_80": {
17691 - "block_sizes": [
17692 - 32,
17693 - 512
17694 - ],
17695 - "loop_orders": [
17696 - [
17697 - 0,
17698 - 1
17699 - ]
17700 - ],
17701 - "flatten_loops": [
17702 - true
17703 - ],
17704 - "l2_groupings": [
17705 - 2
17706 - ],
17707 - "range_unroll_factors": [
17708 - 0
17709 - ],
17710 - "range_warp_specializes": [],
17711 - "range_num_stages": [
17712 - 0
17713 - ],
17714 - "range_multi_buffers": [
17715 - null
17716 - ],
17717 - "range_flattens": [
17718 - null
17719 - ],
17720 - "load_eviction_policies": [
17721 - "",
17722 - "first",
17723 - ""
17724 - ],
17725 - "num_warps": 16,
17726 - "num_stages": 1,
17727 - "indexing": [
17728 - "pointer",
17729 - "tensor_descriptor",
17730 - "pointer",
17731 - "pointer"
17732 - ],
17733 - "pid_type": "flat"
17734 - },
17735 - "intermediate_2048_numtokens_88": {
17736 - "block_sizes": [
17737 - 32,
17738 - 16
17739 - ],
17740 - "loop_orders": [
17741 - [
17742 - 0,
17743 - 1
17744 - ]
17745 - ],
17746 - "flatten_loops": [
17747 - true
17748 - ],
17749 - "l2_groupings": [
17750 - 1
17751 - ],
17752 - "range_unroll_factors": [
17753 - 0
17754 - ],
17755 - "range_warp_specializes": [],
17756 - "range_num_stages": [
17757 - 0
17758 - ],
17759 - "range_multi_buffers": [
17760 - null
17761 - ],
17762 - "range_flattens": [
17763 - null
17764 - ],
17765 - "load_eviction_policies": [
17766 - "",
17767 - "",
17768 - ""
17769 - ],
17770 - "num_warps": 4,
17771 - "num_stages": 2,
17772 - "indexing": [
17773 - "pointer",
17774 - "pointer",
17775 - "pointer",
17776 - "tensor_descriptor"
17777 - ],
17778 - "pid_type": "flat"
17779 - },
17780 - "intermediate_2880_numtokens_88": {
17781 - "block_sizes": [
17782 - 16,
17783 - 128
17784 - ],
17785 - "loop_orders": [
17786 - [
17787 - 0,
17788 - 1
17789 - ]
17790 - ],
17791 - "flatten_loops": [
17792 - true
17793 - ],
17794 - "l2_groupings": [
17795 - 1
17796 - ],
17797 - "range_unroll_factors": [
17798 - 0
17799 - ],
17800 - "range_warp_specializes": [],
17801 - "range_num_stages": [
17802 - 0
17803 - ],
17804 - "range_multi_buffers": [
17805 - null
17806 - ],
17807 - "range_flattens": [
17808 - null
17809 - ],
17810 - "load_eviction_policies": [
17811 - "",
17812 - "",
17813 - ""
17814 - ],
17815 - "num_warps": 4,
17816 - "num_stages": 1,
17817 - "indexing": [
17818 - "tensor_descriptor",
17819 - "pointer",
17820 - "pointer",
17821 - "tensor_descriptor"
17822 - ],
17823 - "pid_type": "flat"
17824 - },
17825 - "intermediate_4096_numtokens_88": {
17826 - "block_sizes": [
17827 - 64,
17828 - 32
17829 - ],
17830 - "loop_orders": [
17831 - [
17832 - 1,
17833 - 0
17834 - ]
17835 - ],
17836 - "flatten_loops": [
17837 - true
17838 - ],
17839 - "l2_groupings": [
17840 - 1
17841 - ],
17842 - "range_unroll_factors": [
17843 - 0
17844 - ],
17845 - "range_warp_specializes": [],
17846 - "range_num_stages": [
17847 - 0
17848 - ],
17849 - "range_multi_buffers": [
17850 - null
17851 - ],
17852 - "range_flattens": [
17853 - null
17854 - ],
17855 - "load_eviction_policies": [
17856 - "",
17857 - "",
17858 - ""
17859 - ],
17860 - "num_warps": 32,
17861 - "num_stages": 1,
17862 - "indexing": [
17863 - "tensor_descriptor",
17864 - "tensor_descriptor",
17865 - "tensor_descriptor",
17866 - "pointer"
17867 - ],
17868 - "pid_type": "flat"
17869 - },
17870 - "intermediate_8192_numtokens_88": {
17871 - "block_sizes": [
17872 - 128,
17873 - 64
17874 - ],
17875 - "loop_orders": [
17876 - [
17877 - 0,
17878 - 1
17879 - ]
17880 - ],
17881 - "flatten_loops": [
17882 - true
17883 - ],
17884 - "l2_groupings": [
17885 - 1
17886 - ],
17887 - "range_unroll_factors": [
17888 - 0
17889 - ],
17890 - "range_warp_specializes": [],
17891 - "range_num_stages": [
17892 - 0
17893 - ],
17894 - "range_multi_buffers": [
17895 - null
17896 - ],
17897 - "range_flattens": [
17898 - null
17899 - ],
17900 - "load_eviction_policies": [
17901 - "first",
17902 - "first",
17903 - "last"
17904 - ],
17905 - "num_warps": 2,
17906 - "num_stages": 1,
17907 - "indexing": [
17908 - "pointer",
17909 - "pointer",
17910 - "pointer",
17911 - "pointer"
17912 - ],
17913 - "pid_type": "flat"
17914 - },
17915 - "intermediate_11008_numtokens_88": {
17916 - "block_sizes": [
17917 - 32,
17918 - 128
17919 - ],
17920 - "loop_orders": [
17921 - [
17922 - 0,
17923 - 1
17924 - ]
17925 - ],
17926 - "flatten_loops": [
17927 - true
17928 - ],
17929 - "l2_groupings": [
17930 - 4
17931 - ],
17932 - "range_unroll_factors": [
17933 - 0
17934 - ],
17935 - "range_warp_specializes": [],
17936 - "range_num_stages": [
17937 - 0
17938 - ],
17939 - "range_multi_buffers": [
17940 - null
17941 - ],
17942 - "range_flattens": [
17943 - null
17944 - ],
17945 - "load_eviction_policies": [
17946 - "",
17947 - "",
17948 - ""
17949 - ],
17950 - "num_warps": 32,
17951 - "num_stages": 1,
17952 - "indexing": [
17953 - "pointer",
17954 - "tensor_descriptor",
17955 - "pointer",
17956 - "pointer"
17957 - ],
17958 - "pid_type": "flat"
17959 - },
17960 - "intermediate_14336_numtokens_88": {
17961 - "block_sizes": [
17962 - 16,
17963 - 128
17964 - ],
17965 - "loop_orders": [
17966 - [
17967 - 0,
17968 - 1
17969 - ]
17970 - ],
17971 - "flatten_loops": [
17972 - true
17973 - ],
17974 - "l2_groupings": [
17975 - 1
17976 - ],
17977 - "range_unroll_factors": [
17978 - 0
17979 - ],
17980 - "range_warp_specializes": [],
17981 - "range_num_stages": [
17982 - 0
17983 - ],
17984 - "range_multi_buffers": [
17985 - null
17986 - ],
17987 - "range_flattens": [
17988 - null
17989 - ],
17990 - "load_eviction_policies": [
17991 - "first",
17992 - "last",
17993 - ""
17994 - ],
17995 - "num_warps": 8,
17996 - "num_stages": 1,
17997 - "indexing": [
17998 - "pointer",
17999 - "pointer",
18000 - "pointer",
18001 - "pointer"
18002 - ],
18003 - "pid_type": "flat"
18004 - },
18005 - "intermediate_2048_numtokens_96": {
18006 - "block_sizes": [
18007 - 128,
18008 - 4
18009 - ],
18010 - "loop_orders": [
18011 - [
18012 - 0,
18013 - 1
18014 - ]
18015 - ],
18016 - "flatten_loops": [
18017 - true
18018 - ],
18019 - "l2_groupings": [
18020 - 2
18021 - ],
18022 - "range_unroll_factors": [
18023 - 0
18024 - ],
18025 - "range_warp_specializes": [],
18026 - "range_num_stages": [
18027 - 0
18028 - ],
18029 - "range_multi_buffers": [
18030 - null
18031 - ],
18032 - "range_flattens": [
18033 - null
18034 - ],
18035 - "load_eviction_policies": [
18036 - "",
18037 - "last",
18038 - ""
18039 - ],
18040 - "num_warps": 4,
18041 - "num_stages": 1,
18042 - "indexing": [
18043 - "pointer",
18044 - "pointer",
18045 - "pointer",
18046 - "pointer"
18047 - ],
18048 - "pid_type": "flat"
18049 - },
18050 - "intermediate_2880_numtokens_96": {
18051 - "block_sizes": [
18052 - 32,
18053 - 128
18054 - ],
18055 - "loop_orders": [
18056 - [
18057 - 0,
18058 - 1
18059 - ]
18060 - ],
18061 - "flatten_loops": [
18062 - true
18063 - ],
18064 - "l2_groupings": [
18065 - 1
18066 - ],
18067 - "range_unroll_factors": [
18068 - 0
18069 - ],
18070 - "range_warp_specializes": [],
18071 - "range_num_stages": [
18072 - 0
18073 - ],
18074 - "range_multi_buffers": [
18075 - null
18076 - ],
18077 - "range_flattens": [
18078 - null
18079 - ],
18080 - "load_eviction_policies": [
18081 - "",
18082 - "",
18083 - ""
18084 - ],
18085 - "num_warps": 16,
18086 - "num_stages": 1,
18087 - "indexing": [
18088 - "pointer",
18089 - "tensor_descriptor",
18090 - "pointer",
18091 - "pointer"
18092 - ],
18093 - "pid_type": "flat"
18094 - },
18095 - "intermediate_4096_numtokens_96": {
18096 - "block_sizes": [
18097 - 16,
18098 - 256
18099 - ],
18100 - "loop_orders": [
18101 - [
18102 - 0,
18103 - 1
18104 - ]
18105 - ],
18106 - "flatten_loops": [
18107 - true
18108 - ],
18109 - "l2_groupings": [
18110 - 2
18111 - ],
18112 - "range_unroll_factors": [
18113 - 0
18114 - ],
18115 - "range_warp_specializes": [],
18116 - "range_num_stages": [
18117 - 0
18118 - ],
18119 - "range_multi_buffers": [
18120 - null
18121 - ],
18122 - "range_flattens": [
18123 - null
18124 - ],
18125 - "load_eviction_policies": [
18126 - "",
18127 - "",
18128 - ""
18129 - ],
18130 - "num_warps": 1,
18131 - "num_stages": 3,
18132 - "indexing": [
18133 - "pointer",
18134 - "pointer",
18135 - "pointer",
18136 - "tensor_descriptor"
18137 - ],
18138 - "pid_type": "flat"
18139 - },
18140 - "intermediate_8192_numtokens_96": {
18141 - "block_sizes": [
18142 - 64,
18143 - 64
18144 - ],
18145 - "loop_orders": [
18146 - [
18147 - 0,
18148 - 1
18149 - ]
18150 - ],
18151 - "flatten_loops": [
18152 - true
18153 - ],
18154 - "l2_groupings": [
18155 - 2
18156 - ],
18157 - "range_unroll_factors": [
18158 - 0
18159 - ],
18160 - "range_warp_specializes": [],
18161 - "range_num_stages": [
18162 - 0
18163 - ],
18164 - "range_multi_buffers": [
18165 - null
18166 - ],
18167 - "range_flattens": [
18168 - null
18169 - ],
18170 - "load_eviction_policies": [
18171 - "first",
18172 - "",
18173 - "last"
18174 - ],
18175 - "num_warps": 1,
18176 - "num_stages": 1,
18177 - "indexing": [
18178 - "pointer",
18179 - "pointer",
18180 - "pointer",
18181 - "pointer"
18182 - ],
18183 - "pid_type": "flat"
18184 - },
18185 - "intermediate_11008_numtokens_96": {
18186 - "block_sizes": [
18187 - 64,
18188 - 256
18189 - ],
18190 - "loop_orders": [
18191 - [
18192 - 0,
18193 - 1
18194 - ]
18195 - ],
18196 - "flatten_loops": [
18197 - true
18198 - ],
18199 - "l2_groupings": [
18200 - 1
18201 - ],
18202 - "range_unroll_factors": [
18203 - 0
18204 - ],
18205 - "range_warp_specializes": [],
18206 - "range_num_stages": [
18207 - 0
18208 - ],
18209 - "range_multi_buffers": [
18210 - null
18211 - ],
18212 - "range_flattens": [
18213 - null
18214 - ],
18215 - "load_eviction_policies": [
18216 - "first",
18217 - "",
18218 - ""
18219 - ],
18220 - "num_warps": 2,
18221 - "num_stages": 2,
18222 - "indexing": [
18223 - "pointer",
18224 - "pointer",
18225 - "pointer",
18226 - "pointer"
18227 - ],
18228 - "pid_type": "flat"
18229 - },
18230 - "intermediate_14336_numtokens_96": {
18231 - "block_sizes": [
18232 - 32,
18233 - 64
18234 - ],
18235 - "loop_orders": [
18236 - [
18237 - 0,
18238 - 1
18239 - ]
18240 - ],
18241 - "flatten_loops": [
18242 - true
18243 - ],
18244 - "l2_groupings": [
18245 - 1
18246 - ],
18247 - "range_unroll_factors": [
18248 - 0
18249 - ],
18250 - "range_warp_specializes": [],
18251 - "range_num_stages": [
18252 - 0
18253 - ],
18254 - "range_multi_buffers": [
18255 - null
18256 - ],
18257 - "range_flattens": [
18258 - null
18259 - ],
18260 - "load_eviction_policies": [
18261 - "",
18262 - "",
18263 - "first"
18264 - ],
18265 - "num_warps": 1,
18266 - "num_stages": 1,
18267 - "indexing": [
18268 - "pointer",
18269 - "pointer",
18270 - "tensor_descriptor",
18271 - "pointer"
18272 - ],
18273 - "pid_type": "flat"
18274 - },
18275 - "intermediate_2048_numtokens_104": {
18276 - "block_sizes": [
18277 - 32,
18278 - 8
18279 - ],
18280 - "loop_orders": [
18281 - [
18282 - 0,
18283 - 1
18284 - ]
18285 - ],
18286 - "flatten_loops": [
18287 - true
18288 - ],
18289 - "l2_groupings": [
18290 - 1
18291 - ],
18292 - "range_unroll_factors": [
18293 - 0
18294 - ],
18295 - "range_warp_specializes": [],
18296 - "range_num_stages": [
18297 - 0
18298 - ],
18299 - "range_multi_buffers": [
18300 - null
18301 - ],
18302 - "range_flattens": [
18303 - null
18304 - ],
18305 - "load_eviction_policies": [
18306 - "",
18307 - "",
18308 - ""
18309 - ],
18310 - "num_warps": 16,
18311 - "num_stages": 2,
18312 - "indexing": [
18313 - "tensor_descriptor",
18314 - "pointer",
18315 - "pointer",
18316 - "pointer"
18317 - ],
18318 - "pid_type": "flat"
18319 - },
18320 - "intermediate_2880_numtokens_104": {
18321 - "block_sizes": [
18322 - 64,
18323 - 64
18324 - ],
18325 - "loop_orders": [
18326 - [
18327 - 0,
18328 - 1
18329 - ]
18330 - ],
18331 - "flatten_loops": [
18332 - true
18333 - ],
18334 - "l2_groupings": [
18335 - 1
18336 - ],
18337 - "range_unroll_factors": [
18338 - 0
18339 - ],
18340 - "range_warp_specializes": [],
18341 - "range_num_stages": [
18342 - 0
18343 - ],
18344 - "range_multi_buffers": [
18345 - null
18346 - ],
18347 - "range_flattens": [
18348 - null
18349 - ],
18350 - "load_eviction_policies": [
18351 - "",
18352 - "",
18353 - ""
18354 - ],
18355 - "num_warps": 32,
18356 - "num_stages": 1,
18357 - "indexing": [
18358 - "pointer",
18359 - "pointer",
18360 - "pointer",
18361 - "pointer"
18362 - ],
18363 - "pid_type": "flat"
18364 - },
18365 - "intermediate_4096_numtokens_104": {
18366 - "block_sizes": [
18367 - 32,
18368 - 32
18369 - ],
18370 - "loop_orders": [
18371 - [
18372 - 0,
18373 - 1
18374 - ]
18375 - ],
18376 - "flatten_loops": [
18377 - false
18378 - ],
18379 - "l2_groupings": [
18380 - 1
18381 - ],
18382 - "range_unroll_factors": [
18383 - 0
18384 - ],
18385 - "range_warp_specializes": [],
18386 - "range_num_stages": [
18387 - 0
18388 - ],
18389 - "range_multi_buffers": [
18390 - null
18391 - ],
18392 - "range_flattens": [
18393 - null
18394 - ],
18395 - "load_eviction_policies": [
18396 - "",
18397 - "",
18398 - ""
18399 - ],
18400 - "num_warps": 8,
18401 - "num_stages": 1,
18402 - "indexing": [
18403 - "pointer",
18404 - "pointer",
18405 - "pointer",
18406 - "pointer"
18407 - ],
18408 - "pid_type": "flat"
18409 - },
18410 - "intermediate_8192_numtokens_104": {
18411 - "block_sizes": [
18412 - 8,
18413 - 8
18414 - ],
18415 - "loop_orders": [
18416 - [
18417 - 0,
18418 - 1
18419 - ]
18420 - ],
18421 - "flatten_loops": [
18422 - true
18423 - ],
18424 - "l2_groupings": [
18425 - 2
18426 - ],
18427 - "range_unroll_factors": [
18428 - 0
18429 - ],
18430 - "range_warp_specializes": [],
18431 - "range_num_stages": [
18432 - 0
18433 - ],
18434 - "range_multi_buffers": [
18435 - null
18436 - ],
18437 - "range_flattens": [
18438 - null
18439 - ],
18440 - "load_eviction_policies": [
18441 - "",
18442 - "",
18443 - ""
18444 - ],
18445 - "num_warps": 8,
18446 - "num_stages": 1,
18447 - "indexing": [
18448 - "pointer",
18449 - "pointer",
18450 - "pointer",
18451 - "tensor_descriptor"
18452 - ],
18453 - "pid_type": "flat"
18454 - },
18455 - "intermediate_11008_numtokens_104": {
18456 - "block_sizes": [
18457 - 128,
18458 - 16
18459 - ],
18460 - "loop_orders": [
18461 - [
18462 - 0,
18463 - 1
18464 - ]
18465 - ],
18466 - "flatten_loops": [
18467 - true
18468 - ],
18469 - "l2_groupings": [
18470 - 1
18471 - ],
18472 - "range_unroll_factors": [
18473 - 0
18474 - ],
18475 - "range_warp_specializes": [],
18476 - "range_num_stages": [
18477 - 0
18478 - ],
18479 - "range_multi_buffers": [
18480 - null
18481 - ],
18482 - "range_flattens": [
18483 - null
18484 - ],
18485 - "load_eviction_policies": [
18486 - "",
18487 - "",
18488 - ""
18489 - ],
18490 - "num_warps": 2,
18491 - "num_stages": 1,
18492 - "indexing": [
18493 - "pointer",
18494 - "tensor_descriptor",
18495 - "pointer",
18496 - "pointer"
18497 - ],
18498 - "pid_type": "flat"
18499 - },
18500 - "intermediate_14336_numtokens_104": {
18501 - "block_sizes": [
18502 - 32,
18503 - 16
18504 - ],
18505 - "loop_orders": [
18506 - [
18507 - 0,
18508 - 1
18509 - ]
18510 - ],
18511 - "flatten_loops": [
18512 - true
18513 - ],
18514 - "l2_groupings": [
18515 - 1
18516 - ],
18517 - "range_unroll_factors": [
18518 - 0
18519 - ],
18520 - "range_warp_specializes": [],
18521 - "range_num_stages": [
18522 - 0
18523 - ],
18524 - "range_multi_buffers": [
18525 - null
18526 - ],
18527 - "range_flattens": [
18528 - null
18529 - ],
18530 - "load_eviction_policies": [
18531 - "",
18532 - "",
18533 - ""
18534 - ],
18535 - "num_warps": 2,
18536 - "num_stages": 1,
18537 - "indexing": [
18538 - "pointer",
18539 - "pointer",
18540 - "pointer",
18541 - "tensor_descriptor"
18542 - ],
18543 - "pid_type": "flat"
18544 - },
18545 - "intermediate_2048_numtokens_112": {
18546 - "block_sizes": [
18547 - 32,
18548 - 1024
18549 - ],
18550 - "loop_orders": [
18551 - [
18552 - 0,
18553 - 1
18554 - ]
18555 - ],
18556 - "flatten_loops": [
18557 - true
18558 - ],
18559 - "l2_groupings": [
18560 - 1
18561 - ],
18562 - "range_unroll_factors": [
18563 - 0
18564 - ],
18565 - "range_warp_specializes": [],
18566 - "range_num_stages": [
18567 - 0
18568 - ],
18569 - "range_multi_buffers": [
18570 - null
18571 - ],
18572 - "range_flattens": [
18573 - null
18574 - ],
18575 - "load_eviction_policies": [
18576 - "first",
18577 - "",
18578 - ""
18579 - ],
18580 - "num_warps": 32,
18581 - "num_stages": 1,
18582 - "indexing": [
18583 - "pointer",
18584 - "tensor_descriptor",
18585 - "pointer",
18586 - "pointer"
18587 - ],
18588 - "pid_type": "flat"
18589 - },
18590 - "intermediate_2880_numtokens_112": {
18591 - "block_sizes": [
18592 - 32,
18593 - 32
18594 - ],
18595 - "loop_orders": [
18596 - [
18597 - 0,
18598 - 1
18599 - ]
18600 - ],
18601 - "flatten_loops": [
18602 - false
18603 - ],
18604 - "l2_groupings": [
18605 - 1
18606 - ],
18607 - "range_unroll_factors": [
18608 - 0
18609 - ],
18610 - "range_warp_specializes": [],
18611 - "range_num_stages": [
18612 - 0
18613 - ],
18614 - "range_multi_buffers": [
18615 - null
18616 - ],
18617 - "range_flattens": [
18618 - null
18619 - ],
18620 - "load_eviction_policies": [
18621 - "",
18622 - "",
18623 - ""
18624 - ],
18625 - "num_warps": 1,
18626 - "num_stages": 1,
18627 - "indexing": [
18628 - "pointer",
18629 - "pointer",
18630 - "pointer",
18631 - "pointer"
18632 - ],
18633 - "pid_type": "flat"
18634 - },
18635 - "intermediate_4096_numtokens_112": {
18636 - "block_sizes": [
18637 - 32,
18638 - 128
18639 - ],
18640 - "loop_orders": [
18641 - [
18642 - 0,
18643 - 1
18644 - ]
18645 - ],
18646 - "flatten_loops": [
18647 - true
18648 - ],
18649 - "l2_groupings": [
18650 - 1
18651 - ],
18652 - "range_unroll_factors": [
18653 - 0
18654 - ],
18655 - "range_warp_specializes": [],
18656 - "range_num_stages": [
18657 - 0
18658 - ],
18659 - "range_multi_buffers": [
18660 - null
18661 - ],
18662 - "range_flattens": [
18663 - null
18664 - ],
18665 - "load_eviction_policies": [
18666 - "last",
18667 - "",
18668 - ""
18669 - ],
18670 - "num_warps": 16,
18671 - "num_stages": 1,
18672 - "indexing": [
18673 - "pointer",
18674 - "pointer",
18675 - "pointer",
18676 - "pointer"
18677 - ],
18678 - "pid_type": "flat"
18679 - },
18680 - "intermediate_8192_numtokens_112": {
18681 - "block_sizes": [
18682 - 32,
18683 - 128
18684 - ],
18685 - "loop_orders": [
18686 - [
18687 - 0,
18688 - 1
18689 - ]
18690 - ],
18691 - "flatten_loops": [
18692 - true
18693 - ],
18694 - "l2_groupings": [
18695 - 1
18696 - ],
18697 - "range_unroll_factors": [
18698 - 0
18699 - ],
18700 - "range_warp_specializes": [],
18701 - "range_num_stages": [
18702 - 0
18703 - ],
18704 - "range_multi_buffers": [
18705 - null
18706 - ],
18707 - "range_flattens": [
18708 - null
18709 - ],
18710 - "load_eviction_policies": [
18711 - "",
18712 - "",
18713 - ""
18714 - ],
18715 - "num_warps": 8,
18716 - "num_stages": 1,
18717 - "indexing": [
18718 - "pointer",
18719 - "pointer",
18720 - "pointer",
18721 - "tensor_descriptor"
18722 - ],
18723 - "pid_type": "flat"
18724 - },
18725 - "intermediate_11008_numtokens_112": {
18726 - "block_sizes": [
18727 - 16,
18728 - 64
18729 - ],
18730 - "loop_orders": [
18731 - [
18732 - 0,
18733 - 1
18734 - ]
18735 - ],
18736 - "flatten_loops": [
18737 - true
18738 - ],
18739 - "l2_groupings": [
18740 - 1
18741 - ],
18742 - "range_unroll_factors": [
18743 - 0
18744 - ],
18745 - "range_warp_specializes": [],
18746 - "range_num_stages": [
18747 - 0
18748 - ],
18749 - "range_multi_buffers": [
18750 - null
18751 - ],
18752 - "range_flattens": [
18753 - null
18754 - ],
18755 - "load_eviction_policies": [
18756 - "",
18757 - "",
18758 - ""
18759 - ],
18760 - "num_warps": 8,
18761 - "num_stages": 1,
18762 - "indexing": [
18763 - "pointer",
18764 - "pointer",
18765 - "pointer",
18766 - "pointer"
18767 - ],
18768 - "pid_type": "flat"
18769 - },
18770 - "intermediate_14336_numtokens_112": {
18771 - "block_sizes": [
18772 - 32,
18773 - 8
18774 - ],
18775 - "loop_orders": [
18776 - [
18777 - 0,
18778 - 1
18779 - ]
18780 - ],
18781 - "flatten_loops": [
18782 - true
18783 - ],
18784 - "l2_groupings": [
18785 - 2
18786 - ],
18787 - "range_unroll_factors": [
18788 - 0
18789 - ],
18790 - "range_warp_specializes": [],
18791 - "range_num_stages": [
18792 - 0
18793 - ],
18794 - "range_multi_buffers": [
18795 - null
18796 - ],
18797 - "range_flattens": [
18798 - null
18799 - ],
18800 - "load_eviction_policies": [
18801 - "",
18802 - "",
18803 - ""
18804 - ],
18805 - "num_warps": 1,
18806 - "num_stages": 2,
18807 - "indexing": [
18808 - "pointer",
18809 - "pointer",
18810 - "pointer",
18811 - "pointer"
18812 - ],
18813 - "pid_type": "flat"
18814 - },
18815 - "intermediate_2048_numtokens_120": {
18816 - "block_sizes": [
18817 - 32,
18818 - 64
18819 - ],
18820 - "loop_orders": [
18821 - [
18822 - 0,
18823 - 1
18824 - ]
18825 - ],
18826 - "flatten_loops": [
18827 - true
18828 - ],
18829 - "l2_groupings": [
18830 - 1
18831 - ],
18832 - "range_unroll_factors": [
18833 - 0
18834 - ],
18835 - "range_warp_specializes": [],
18836 - "range_num_stages": [
18837 - 0
18838 - ],
18839 - "range_multi_buffers": [
18840 - null
18841 - ],
18842 - "range_flattens": [
18843 - null
18844 - ],
18845 - "load_eviction_policies": [
18846 - "",
18847 - "",
18848 - ""
18849 - ],
18850 - "num_warps": 2,
18851 - "num_stages": 1,
18852 - "indexing": [
18853 - "pointer",
18854 - "pointer",
18855 - "pointer",
18856 - "pointer"
18857 - ],
18858 - "pid_type": "flat"
18859 - },

This file is too large to show in full.

vllm/kernels/helion/utils.py
+45 -5
@@ -8,6 +8,44 @@ from vllm.platforms import current_platform
8
9 logger = logging.getLogger(__name__)
10
11 +# Maps known variant GPU names (after lowercase/underscore normalization)
12 +# to their canonical form.
13 +#
14 +# Names that are already canonical after normalization are NOT listed here.
15 +# For example, "NVIDIA H200" normalizes to "nvidia_h200" which needs no
16 +# further mapping, and AMD ROCm names like "AMD_Instinct_MI300X" come from
17 +# a controlled lookup table in rocm.py and normalize cleanly to
18 +# "amd_instinct_mi300x". Only names with variant suffixes (form factor,
19 +# memory size, memory type, etc.) that should be stripped need entries.
20 +#
21 +# To add a new GPU variant: run `canonicalize_gpu_name()` without the alias
22 +# to see the normalized name, then add a mapping here if it contains variant
23 +# suffixes that should be stripped (e.g. Blackwell/Rubin variants).
24 +_GPU_NAME_ALIASES: dict[str, str] = {
25 + # H100 variants
26 + "nvidia_h100_pcie": "nvidia_h100",
27 + "nvidia_h100_sxm5": "nvidia_h100",
28 + "nvidia_h100_80gb_hbm3": "nvidia_h100",
29 + "nvidia_h100_nvl": "nvidia_h100",
30 + # H200 variants
31 + "nvidia_h200_nvl": "nvidia_h200",
32 + "nvidia_h200_141gb_hbm3e": "nvidia_h200",
33 + # A100 variants
34 + "nvidia_a100_sxm4_80gb": "nvidia_a100",
35 + "nvidia_a100_sxm4_40gb": "nvidia_a100",
36 + "nvidia_a100_pcie_80gb": "nvidia_a100",
37 + "nvidia_a100_pcie_40gb": "nvidia_a100",
38 + "nvidia_a100_80gb_pcie": "nvidia_a100",
39 + # V100 variants (Tesla-branded)
40 + "tesla_v100_sxm2_32gb": "tesla_v100",
41 + "tesla_v100_sxm2_16gb": "tesla_v100",
42 + "tesla_v100_pcie_32gb": "tesla_v100",
43 + "tesla_v100_pcie_16gb": "tesla_v100",
44 + # AMD ROCm variants (from _ROCM_DEVICE_ID_NAME_MAP in rocm.py)
45 + "amd_instinct_mi300x_hf": "amd_instinct_mi300x",
46 + # ADD MORE HERE
47 +}
48 +
49
50 def get_gpu_name(device_id: int | None = None) -> str:
51 if device_id is None:
@@ -23,17 +61,19 @@ def canonicalize_gpu_name(name: str) -> str:
61 """
62 Canonicalize GPU name for use as a platform identifier.
63
26 - Converts to lowercase and replaces spaces and hyphens with underscores.
27 - e.g., "NVIDIA A100-SXM4-80GB" -> "nvidia_a100_sxm4_80gb"
28 - "AMD_Instinct_MI300X" -> "amd_instinct_mi300x"
29 -
30 - Raises ValueError if name is empty.
64 + Converts to lowercase, replaces spaces and hyphens with underscores,
65 + and maps known variant names to their canonical form via _GPU_NAME_ALIASES.
66 + e.g., "NVIDIA H100 80GB HBM3" -> "nvidia_h100"
67 + "NVIDIA A100-SXM4-80GB" -> "nvidia_a100"
68 + "AMD Instinct MI300X" -> "amd_instinct_mi300x"
69 """
70 if not name or not name.strip():
71 raise ValueError("GPU name cannot be empty")
72 name = name.lower()
73 name = name.replace(" ", "_")
74 name = name.replace("-", "_")
75 + if name in _GPU_NAME_ALIASES:
76 + return _GPU_NAME_ALIASES[name]
77 return name
78
79