[rahulbatra85]

Motivation

--Currently, only tensors are accepted for building rotating buffers pool. Allow list/dict/tuple of tensors in generate_rotating_buffers_pool
--Separate out RotatingBuffer and do_bench_aiter_triton into it's own file
--pool target_mb is not set based on the arch/device

PR Self-Review

What behavior does this PR change, and why is the existing behavior insufficient?
-separate rotating buffers and do_bench_aiter_triton into it's own file
-new API using class called RotatingBufferBench
-use defaults for target_mb based on arch.
-determine max_gpu_mb based on free device memory available when creating the pool

What did you validate, and what remains unverified? Include commands, hardware/configuration and results.
-Just tried the change with benchmark that needs it.

What behavior would each changed op test fail on before this PR, and why is that coverage not already provided by an existing canonical test?
--This adds functionality for Triton benchmarks. There are no test for benchmarks.

For every changed op_tests/ file, what behavior does it cover and why is it not redundant? Write None if no op test changed.
-This adds functionality for Triton benchmarks. There are no test for benchmarks.

[github-actions]

🏷️ CI Guide

Runs automatically on every PR:

• ✅ Pre-checks (submodule verification, code formatting)
• ✅ Aiter op tests (gfx942 + gfx950)
• ✅ Triton tests on MI35X (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
ci:gfx1250-ffm-triton	Run the five-shard gfx1250 FFM Triton test suite
ci:triton-300x	Run an additional Triton test job on MI300X in PRs (added automatically when gfx942 configs change); main branch always runs both MI35X and MI300X
ci:triton-355	Run the full Triton test suite on MI35X, not only the tests the change affects
ci:windows	Windows JIT op smoke test on the gfx1201 ROCm runner
ci:multi-gpu	Aiter multi-GPU tests on the 8-GPU runner
ci:sglang	SGLang integration tests: DeepSeek-R1-MXFP4 accuracy, Qwen 3.5 accuracy
ci:atom	ATOM benchmark: DeepSeek-R1-0528, GPT-OSS-120B
ci:atom_full	ATOM accuracy suite for PR and main models from ATOM models_accuracy.json
ci:vllm	vLLM benchmark: DeepSeek-V4-Pro, MiniMax-M3-MXFP4
ci:all	All standard extended tests (excludes ci:atom_full)

Only add ci:atom_full for FlyDSL or Triton upgrades.
Add labels via the sidebar or gh pr edit 7121 --add-label <label>

One backend per PR:
A PR changes one kernel backend: [Triton/Gluon] (Triton and Gluon count as one), [HIP], [ASM], [CK], [OPUS] or [FlyDSL]. If the title ends up with two backend tags, split the PR -- as stacked pull requests when one part cannot merge without the other.

PR title tags & labels:
Component tags ([Triton/Gluon], [HIP], [CK], [ASM], ...) are added to the PR title and as PR labels automatically from the changed files and re-synced on every push — change-type tags like [fix]/[Perf], op tags like [MLA], and human labels (ci:*) are left untouched. Add the no-auto-title label to stop the title rewrites; labels stay in sync either way.

[Copilot]

🟡 Changes recommended

The sizing loop retains its last tensor sample during pool allocation, increasing peak GPU memory and potentially causing an otherwise fitting pool to run out of memory.

1 open issue

• Release sizing tensor before building pool · New

What changed in this PR

This PR extends the Triton/Gluon benchmark helper to size rotating buffer pools from nested tensor inputs and exclude selected fields from sizing.

Changes:

• Adds exclude_from_sizing to omit named fields from the memory estimate.
• Recursively counts tensors in dictionaries, lists, and tuples while preserving storage deduplication.

File	Description
op_tests/​op_benchmarks/​triton/​utils/​benchmark_utils.py	Adds field exclusions and recursive tensor counting for rotating pool sizing.

_{Review effort: Balanced}

[aiter-bot]

NOT READY FOR HUMAN REVIEW

PR Self-Review

Revision 88f4014d83b0 · mode shadow

Waiting for the PR author. The readiness report will appear after the self-review answers are complete for this revision.

Questions to answer in the PR description

1. What behavior does this PR change, and why is the existing behavior insufficient?
2. What did you validate, and what remains unverified? Include commands, hardware/configuration and results.
3. What behavior would each changed op test fail on before this PR, and why is that coverage not already provided by an existing canonical test?
4. For every changed op_tests/ file, what behavior does it cover and why is it not redundant? Write None if no op test changed.

Edit the PR Self-Review section in the PR description. Include the current revision SHA; this comment refreshes automatically.

Missing or invalid answers

• FAIL — missing ### Reviewed revision in the PR description
• FAIL — missing ### Behavior and intent in the PR description
• FAIL — missing ### Validation and unverified scope in the PR description
• FAIL — missing ### PR-specific question in the PR description
• FAIL — missing ### Changed op tests in the PR description
• FAIL — Reviewed revision must contain at least the first 7 characters of the current head SHA
• FAIL — Behavior and intent has no substantive answer
• FAIL — Validation and unverified scope has no substantive answer
• FAIL — PR-specific question has no substantive answer
• FAIL — changed op test not explained in Changed op tests: op_tests/op_benchmarks/triton/AGENTS.md
• FAIL — changed op test not explained in Changed op tests: op_tests/op_benchmarks/triton/README.md
• FAIL — changed op test not explained in Changed op tests: op_tests/op_benchmarks/triton/bench_all_reduce.py
• FAIL — changed op test not explained in Changed op tests: op_tests/op_benchmarks/triton/bench_all_reduce_rmsnorm_quant.py
• FAIL — changed op test not explained in Changed op tests: op_tests/op_benchmarks/triton/bench_batched_gemm_a16wfp4.py
• FAIL — changed op test not explained in Changed op tests: op_tests/op_benchmarks/triton/bench_batched_gemm_a8w8.py
• FAIL — changed op test not explained in Changed op tests: op_tests/op_benchmarks/triton/bench_batched_gemm_a8w8_a_per_token_group_prequant_w_per_batched_tensor_quant.py
• FAIL — changed op test not explained in Changed op tests: op_tests/op_benchmarks/triton/bench_batched_gemm_afp4wfp4.py
• FAIL — 48 additional changed op test(s) are not explained
• FAIL — Changed op tests must explain why each changed op test is necessary and non-redundant

_{Shadow mode reports what would block but keeps the status successful. Set AITER_PR_SELF_REVIEW_MODE=enforce only after reviewing real false positives.}

[github-actions]

Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit

ruff

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_a8w8_blockscale_a16w16.py

Lines 1 to 25 in cb4a98e

	import functools
	import torch
	from aiter.ops.triton.gemm.fused.fused_gemm_a8w8_blockscale_a16w16 import (
	fused_gemm_a8w8_blockscale_a16w16,
	)
	from op_tests.op_benchmarks.triton.bench_fused_gemm_commons import (
	metric_to_scalar,
	parse_fused_args,
	run_fused_benchmark,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8_blockscale import (
	generate_gemm_a8w8_blockscale_inputs,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16 import (
	generate_gemm_a16w16_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_a8w8_blockscale_mul_add.py

Lines 1 to 23 in cb4a98e

	import functools
	import torch
	from aiter.ops.triton.gemm.fused.fused_gemm_a8w8_blockscale_mul_add import (
	fused_gemm_a8w8_blockscale_mul_add,
	)
	from op_tests.op_benchmarks.triton.bench_fused_gemm_commons import (
	make_mul_add_ab,
	metric_to_scalar,
	parse_fused_args,
	run_fused_benchmark,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8_blockscale import (
	generate_gemm_a8w8_blockscale_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_afp4wfp4_a16w16.py

Lines 1 to 25 in cb4a98e

	import functools
	import torch
	from aiter.ops.triton.gemm.fused.fused_gemm_afp4wfp4_a16w16 import (
	fused_gemm_afp4wfp4_a16w16,
	)
	from op_tests.op_benchmarks.triton.bench_fused_gemm_commons import (
	metric_to_scalar,
	parse_fused_args,
	run_fused_benchmark,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16 import (
	generate_gemm_a16w16_inputs,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_afp4wfp4 import (
	generate_gemm_afp4wfp4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_afp4wfp4_mul_add.py

Lines 1 to 23 in cb4a98e

	import functools
	import torch
	from aiter.ops.triton.gemm.fused.fused_gemm_afp4wfp4_mul_add import (
	fused_gemm_afp4wfp4_mul_add,
	)
	from op_tests.op_benchmarks.triton.bench_fused_gemm_commons import (
	make_mul_add_ab,
	metric_to_scalar,
	parse_fused_args,
	run_fused_benchmark,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_afp4wfp4 import (
	generate_gemm_afp4wfp4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_commons.py

Lines 1 to 16 in cb4a98e

	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.fusions.test_fused_mul_add import (
	generate_fused_mul_add_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_indexed_gated_residual.py

Lines 9 to 31 in cb4a98e

	import functools
	import sys
	import torch
	import triton
	from aiter.ops.triton.fusions.fused_indexed_gated_residual import (
	fused_indexed_gated_residual,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.fusions.test_fused_indexed_gated_residual import make

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_kda_decode.py

Lines 17 to 38 in cb4a98e

	import functools
	import torch
	import triton
	import triton.language as tl
	from einops import rearrange
	from aiter.ops.triton._triton_kernels.gated_delta_net.decode.fused_sigmoid_gating_recurrent import (
	fused_sigmoid_gating_delta_rule_update,
	)
	from aiter.ops.triton.gated_delta_net.causal_conv1d_decode import (
	causal_conv1d_update_split_qkv,
	)
	from aiter.ops.triton.gated_delta_net.fused_kda_decode import fused_kda_decode
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_mla_prefill_qkv_fp8.py

Lines 6 to 32 in cb4a98e

	import functools
	import sys
	import torch
	import triton
	from aiter import dtypes
	from aiter.ops.triton.fusions.fused_mla_prefill_qkv_fp8 import (
	fused_mla_prefill_qkv_fp8,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.fusions.test_fused_mla_prefill_qkv_fp8 import (
	generate_inputs,
	torch_mla_prefill_qkv_fp8_ref,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_qk_cat_and_cache_mla.py

Lines 30 to 50 in cb4a98e

	import functools
	import torch
	import triton
	from aiter.ops.triton.fusions.fused_kv_cache import (
	fused_qk_cat_and_cache_mla,
	fused_qk_rope_cat_and_cache_mla,
	)
	from aiter.ops.triton.utils._triton import arch_info
	from aiter.ops.triton.utils.types import e4m3_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_qk_norm_rope_gate_fp8_quant.py

Lines 14 to 33 in cb4a98e

	import functools
	import torch
	import triton
	from aiter.ops.triton.rope.fused_qk_norm_rope_gate_fp8_quant import (
	FP8_DTYPE,
	MAX_SEQUENCES,
	fused_qk_norm_rope_gate_fp8_quant,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_qk_rope_reshape_and_cache.py

Lines 24 to 42 in cb4a98e

	import argparse
	import functools
	import torch
	import triton
	from aiter.ops.triton.fusions.fused_kv_cache import fused_qk_rope_reshape_and_cache
	from aiter.ops.triton.utils._triton import arch_info
	from aiter.ops.triton.utils.types import e4m3_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_reduce_qk_norm_rope_swa_write.py

Lines 9 to 34 in cb4a98e

	import argparse
	import functools
	import torch
	import triton
	from aiter.ops.triton.fusions.fused_reduce_qk_norm_rope_swa_write import (
	fused_reduce_qk_norm_rope_swa_write,
	)
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.fusions.test_fused_reduce_qk_norm_rope_swa_write import (
	_build_cos_sin,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_rms_gated_mxfp4_quant.py

Lines 1 to 18 in cb4a98e

	import functools
	import sys
	import torch
	import torch.nn.functional as F
	import triton
	from aiter.ops.triton.quant import dynamic_mxfp4_quant, fused_rms_gated_mxfp4_quant
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_rmsnorm_add.py

Lines 1 to 22 in cb4a98e

	import functools
	import triton
	from aiter.ops.triton.normalization.fused_rmsnorm_add import fused_rmsnorm_add
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.normalization.test_fused_rmsnorm_add import (
	generate_fused_rmsnorm_add_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_rmsnorm_add_rmsnorm.py

Lines 10 to 30 in cb4a98e

	import functools
	import torch
	import triton
	from aiter.ops.triton.normalization.fused_rmsnorm_add_rmsnorm import (
	fused_rmsnorm_add_rmsnorm,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.normalization.test_fused_rmsnorm_add_rmsnorm import (
	generate_fused_rmsnorm_add_rmsnorm_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_sigmoid_mul.py

Lines 8 to 30 in cb4a98e

	import functools
	import sys
	import torch
	import triton
	from aiter.ops.triton.fusions.fused_sigmoid_mul import fused_sigmoid_mul
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.fusions.test_fused_sigmoid_mul import (
	generate_fused_sigmoid_mul_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w16.py

Lines 1 to 27 in cb4a98e

	import functools
	import math
	import torch
	import triton
	from aiter.ops.triton.gemm.basic.gemm_a16w16 import _is_gluon_available, gemm_a16w16
	from aiter.ops.triton.gemm.basic.gemm_a16w16_atomic import gemm_a16w16_atomic
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16 import (
	generate_gemm_a16w16_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w16_gated.py

Lines 1 to 26 in cb4a98e

	import functools
	import math
	import torch
	import triton
	from aiter.ops.triton.gemm.basic.gemm_a16w16_gated import gemm_a16w16_gated
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16_gated import (
	generate_gemm_a16w16_gated_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w16_gating.py

Lines 1 to 26 in cb4a98e

	import functools
	import math
	import sys
	import matplotlib.pyplot as plt
	import torch
	import triton
	from aiter.ops.triton.gemm.basic.gemm_a16w16_gated import gemm_a16w16_gated
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_shape_benchmark_object,
	model_benchmark_shapes,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16_gated import (
	generate_gemm_a16w16_gated_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w4.py

Lines 1 to 27 in cb4a98e

	import functools
	import math
	import torch
	import triton
	from aiter import logger
	from aiter.ops.triton.gemm.basic.gemm_a16w4 import gemm_a16w4
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w4 import (
	generate_gemm_a16w4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w8_blockscale.py

Lines 1 to 28 in cb4a98e

	import functools
	import math
	import torch
	import triton
	from aiter.ops.triton.gemm.basic.gemm_a16w8_blockscale import (
	gemm_a16w8_blockscale,
	gemm_a16w8_blockscale_preshuffle,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w8_blockscale import (
	generate_gemm_a16w8_blockscale_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a8w8.py

Lines 1 to 29 in cb4a98e

	import functools
	import math
	import sys
	from collections.abc import Callable
	import triton
	from aiter.ops.triton.gemm.basic.gemm_a8w8 import gemm_a8w8 as triton_gemm_a8w8
	from aiter.ops.triton.gemm.basic.gemm_a8w8 import gemm_a8w8_preshuffle
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8 import (
	generate_gemm_a8w8_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a8w8_blockscale.py

Lines 1 to 34 in cb4a98e

	import functools
	import math
	from collections.abc import Callable
	import torch
	import triton
	from aiter.ops.triton.gemm.basic.gemm_a8w8_blockscale import (
	gemm_a8w8_blockscale as triton_gemm_a8w8_blockscale,
	)
	from aiter.ops.triton.gemm.basic.gemm_a8w8_blockscale import (
	gemm_a8w8_blockscale_preshuffle as triton_gemm_a8w8_blockscale_preshuffle,
	)
	from aiter.test_common import checkAllclose
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8_blockscale import (
	generate_gemm_a8w8_blockscale_inputs,
	run_torch,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a8w8_per_token_scale.py

Lines 1 to 28 in cb4a98e

	import functools
	import math
	import torch
	import triton
	from aiter.ops.triton.gemm.basic.gemm_a8w8_per_token_scale import (
	gemm_a8w8_per_token_scale,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8_per_token_scale import (
	generate_gemm_a8w8_per_token_scale_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a8wfp4.py

Lines 1 to 30 in cb4a98e

	import functools
	import math
	import sys
	import torch
	import triton
	from aiter.ops.triton.gemm.basic.gemm_a8wfp4 import (
	gemm_a8wfp4,
	)
	from aiter.ops.triton.utils.types import get_fp8_dtypes
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8wfp4 import (
	generate_gemm_a8wfp4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_afp4wfp4.py

Lines 1 to 31 in cb4a98e

	import functools
	import math
	import torch
	import triton
	from aiter.ops.triton.gemm.basic.gemm_afp4wfp4 import (
	gemm_afp4wfp4 as triton_gemm_afp4wfp4,
	)
	from aiter.ops.triton.gemm.basic.gemm_afp4wfp4 import (
	gemm_afp4wfp4_preshuffle,
	)
	from aiter.ops.triton.utils._triton import arch_info
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_afp4wfp4 import (
	generate_gemm_afp4wfp4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_afp8wfp8.py

Lines 11 to 40 in cb4a98e

	import csv
	import functools
	import math
	import warnings
	from pathlib import Path
	import torch
	import triton
	from aiter.ops.shuffle import shuffle_weight
	from aiter.ops.triton.gemm.basic.gemm_afp8wfp8 import (
	gemm_afp8wfp8,
	gemm_afp8wfp8_preshuffle,
	)
	from aiter.ops.triton.utils._triton import arch_info
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gmm.py

Lines 9 to 64 in cb4a98e

	import argparse
	import functools
	import logging
	# PyTorch
	import torch
	# Triton
	import triton
	# AITER: Triton kernel wrappers
	from aiter.ops.triton.gemm.grouped.gmm import (
	gmm as triton_gmm,
	)
	from aiter.ops.triton.gemm.grouped.gmm import (
	nptgmm as triton_nptgmm,
	)
	from aiter.ops.triton.gemm.grouped.gmm import (
	ptgmm as triton_ptgmm,
	)
	# AITER: GMM defaults and utility functions
	from aiter.ops.triton.utils.gmm_common import (
	DTYPE,
	DTYPE_STR,
	GROUP_SIZES_DTYPE,
	GROUP_SIZES_DTYPE_STR,
	NUM_GROUP_SIZES,
	RNG_SEED,
	SUPPORTED_DTYPES_STR,
	SUPPORTED_GROUP_SIZES_DTYPES_STR,
	TRANS_LHS,
	TRANS_RHS,
	dtype_from_str,
	gen_gmm_input,
	gen_gmm_output,
	gen_multiple_group_sizes,
	gen_tgmm_bias_grad,
	gen_tgmm_input,
	gen_tgmm_output,
	group_sizes_dtype_from_str,
	str_from_dtype,
	str_from_group_sizes_dtype,
	)
	# AITER: shared argparse helper
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	positive_int,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_layernorm.py

Lines 1 to 27 in cb4a98e

	import functools
	import torch
	import triton
	from aiter.ops.triton.normalization.norm import (
	layer_norm,
	layernorm2d_fwd_with_add,
	layernorm2d_fwd_with_add_dynamicquant,
	layernorm2d_fwd_with_add_smoothquant,
	layernorm2d_fwd_with_dynamicquant,
	layernorm2d_fwd_with_smoothquant,
	)
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_mla.py

Lines 6 to 24 in cb4a98e

	import functools
	import torch
	import triton
	from aiter.ops.triton.attention.mla import mla_decode_fwd, mla_prefill_fwd
	from aiter.ops.triton.utils._triton import arch_info
	from aiter.ops.triton.utils.types import e4m3_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.attention.test_mla import input_helper, shuffle_kv_buffer

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_mla_decode_rope.py

Lines 1 to 29 in cb4a98e

	import argparse
	import functools
	import itertools
	import sys
	import torch
	import triton
	from aiter.ops.triton.attention.mla_decode_rope import (
	decode_attention_fwd_grouped_rope,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_configs,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.attention.test_mla_decode_rope import (
	input_helper,
	ref_preprocess,
	)
	from op_tests.triton_tests.utils.rotary_embedding import DeepseekScalingRotaryEmbedding

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_rmsnorm.py

Lines 1 to 25 in cb4a98e

	import functools
	import math
	import torch
	import triton
	from aiter.ops.triton.normalization.rmsnorm import rms_norm
	from aiter.ops.triton.quant.fused_mxfp4_quant import fused_rms_mxfp4_quant
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_configs,
	print_vgpr,
	)
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.normalization.test_rmsnorm import (
	generate_rmsnorm_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_rope.py

Lines 1 to 40 in cb4a98e

	import functools
	import torch
	from aiter.ops.triton.rope.rope import (
	RotateStyle,
	rope_bwd,
	rope_cached_bwd,
	rope_cached_fwd,
	rope_cached_fwd_inplace,
	rope_cached_positions_bwd,
	rope_cached_positions_fwd,
	rope_cached_positions_fwd_inplace,
	rope_cached_positions_offsets_bwd,
	rope_cached_positions_offsets_fwd,
	rope_cached_positions_offsets_fwd_inplace,
	rope_cached_thd_positions_2c_bwd,
	rope_cached_thd_positions_2c_fwd,
	rope_cached_thd_positions_2c_fwd_inplace,
	rope_cached_thd_positions_offsets_2c_bwd,
	# rope_fwd_2d,
	# rope_fwd_2d_inplace,
	rope_cached_thd_positions_offsets_2c_fwd,
	rope_cached_thd_positions_offsets_2c_fwd_inplace,
	rope_fwd,
	rope_fwd_inplace,
	rope_thd_bwd,
	rope_thd_fwd,
	rope_thd_fwd_inplace,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_model_configs
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.triton_tests.rope.test_rope import generate_rope_inputs

[github-actions]

Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit

ruff

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_a8w8_blockscale_mul_add.py

Lines 1 to 23 in 5332c56

	import functools
	import torch
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.fused.fused_gemm_a8w8_blockscale_mul_add import (
	fused_gemm_a8w8_blockscale_mul_add,
	)
	from op_tests.op_benchmarks.triton.bench_fused_gemm_commons import (
	make_mul_add_ab,
	metric_to_scalar,
	parse_fused_args,
	run_fused_benchmark,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8_blockscale import (
	generate_gemm_a8w8_blockscale_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_afp4wfp4_a16w16.py

Lines 1 to 25 in 5332c56

	import functools
	import torch
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.fused.fused_gemm_afp4wfp4_a16w16 import (
	fused_gemm_afp4wfp4_a16w16,
	)
	from op_tests.op_benchmarks.triton.bench_fused_gemm_commons import (
	metric_to_scalar,
	parse_fused_args,
	run_fused_benchmark,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16 import (
	generate_gemm_a16w16_inputs,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_afp4wfp4 import (
	generate_gemm_afp4wfp4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_afp4wfp4_mul_add.py

Lines 1 to 23 in 5332c56

	import functools
	import torch
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.fused.fused_gemm_afp4wfp4_mul_add import (
	fused_gemm_afp4wfp4_mul_add,
	)
	from op_tests.op_benchmarks.triton.bench_fused_gemm_commons import (
	make_mul_add_ab,
	metric_to_scalar,
	parse_fused_args,
	run_fused_benchmark,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_afp4wfp4 import (
	generate_gemm_afp4wfp4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_gemm_commons.py

Lines 1 to 16 in 5332c56

	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.triton_tests.fusions.test_fused_mul_add import (
	generate_fused_mul_add_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_indexed_gated_residual.py

Lines 9 to 31 in 5332c56

	import functools
	import sys
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.fusions.fused_indexed_gated_residual import (
	fused_indexed_gated_residual,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.fusions.test_fused_indexed_gated_residual import make

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_kda_decode.py

Lines 17 to 38 in 5332c56

	import functools
	import torch
	import triton
	import triton.language as tl
	from einops import rearrange
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton._triton_kernels.gated_delta_net.decode.fused_sigmoid_gating_recurrent import (
	fused_sigmoid_gating_delta_rule_update,
	)
	from aiter.ops.triton.gated_delta_net.causal_conv1d_decode import (
	causal_conv1d_update_split_qkv,
	)
	from aiter.ops.triton.gated_delta_net.fused_kda_decode import fused_kda_decode
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_mla_prefill_qkv_fp8.py

Lines 6 to 32 in 5332c56

	import functools
	import sys
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter import dtypes
	from aiter.ops.triton.fusions.fused_mla_prefill_qkv_fp8 import (
	fused_mla_prefill_qkv_fp8,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.fusions.test_fused_mla_prefill_qkv_fp8 import (
	generate_inputs,
	torch_mla_prefill_qkv_fp8_ref,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_qk_cat_and_cache_mla.py

Lines 30 to 50 in 5332c56

	import functools
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.fusions.fused_kv_cache import (
	fused_qk_cat_and_cache_mla,
	fused_qk_rope_cat_and_cache_mla,
	)
	from aiter.ops.triton.utils._triton import arch_info
	from aiter.ops.triton.utils.types import e4m3_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_qk_concat.py

Lines 6 to 30 in 5332c56

	import functools
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.fusions.fused_qk_concat import fused_qk_cat, fused_qk_rope_cat
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	positive_int,
	pow2_int,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.fusions.test_fused_qk_concat import (
	generate_qk_inputs,
	generate_rope_cached_freqs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_qk_norm_rope_gate_fp8_quant.py

Lines 14 to 33 in 5332c56

	import functools
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.rope.fused_qk_norm_rope_gate_fp8_quant import (
	FP8_DTYPE,
	MAX_SEQUENCES,
	fused_qk_norm_rope_gate_fp8_quant,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_qk_rope_reshape_and_cache.py

Lines 24 to 42 in 5332c56

	import argparse
	import functools
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.fusions.fused_kv_cache import fused_qk_rope_reshape_and_cache
	from aiter.ops.triton.utils._triton import arch_info
	from aiter.ops.triton.utils.types import e4m3_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_reduce_qk_norm_rope_swa_write.py

Lines 9 to 34 in 5332c56

	import argparse
	import functools
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.fusions.fused_reduce_qk_norm_rope_swa_write import (
	fused_reduce_qk_norm_rope_swa_write,
	)
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.fusions.test_fused_reduce_qk_norm_rope_swa_write import (
	_build_cos_sin,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_rmsnorm_add.py

Lines 1 to 22 in 5332c56

	import functools
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.normalization.fused_rmsnorm_add import fused_rmsnorm_add
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.normalization.test_fused_rmsnorm_add import (
	generate_fused_rmsnorm_add_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_rmsnorm_add_rmsnorm.py

Lines 10 to 30 in 5332c56

	import functools
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.normalization.fused_rmsnorm_add_rmsnorm import (
	fused_rmsnorm_add_rmsnorm,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext
	from op_tests.triton_tests.normalization.test_fused_rmsnorm_add_rmsnorm import (
	generate_fused_rmsnorm_add_rmsnorm_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_fused_sigmoid_mul.py

Lines 8 to 30 in 5332c56

	import functools
	import sys
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.fusions.fused_sigmoid_mul import fused_sigmoid_mul
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)
	from op_tests.triton_tests.fusions.test_fused_sigmoid_mul import (
	generate_fused_sigmoid_mul_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w16.py

Lines 1 to 27 in 5332c56

	import functools
	import math
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_a16w16 import _is_gluon_available, gemm_a16w16
	from aiter.ops.triton.gemm.basic.gemm_a16w16_atomic import gemm_a16w16_atomic
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16 import (
	generate_gemm_a16w16_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w16_gated.py

Lines 1 to 26 in 5332c56

	import functools
	import math
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_a16w16_gated import gemm_a16w16_gated
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16_gated import (
	generate_gemm_a16w16_gated_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w16_gating.py

Lines 1 to 26 in 5332c56

	import functools
	import math
	import sys
	import matplotlib.pyplot as plt
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_a16w16_gated import gemm_a16w16_gated
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_shape_benchmark_object,
	model_benchmark_shapes,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w16_gated import (
	generate_gemm_a16w16_gated_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w4.py

Lines 1 to 27 in 5332c56

	import functools
	import math
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter import logger
	from aiter.ops.triton.gemm.basic.gemm_a16w4 import gemm_a16w4
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w4 import (
	generate_gemm_a16w4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a16w8_blockscale.py

Lines 1 to 28 in 5332c56

	import functools
	import math
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_a16w8_blockscale import (
	gemm_a16w8_blockscale,
	gemm_a16w8_blockscale_preshuffle,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a16w8_blockscale import (
	generate_gemm_a16w8_blockscale_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a8w8.py

Lines 1 to 29 in 5332c56

	import functools
	import math
	import sys
	from collections.abc import Callable
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_a8w8 import gemm_a8w8 as triton_gemm_a8w8
	from aiter.ops.triton.gemm.basic.gemm_a8w8 import gemm_a8w8_preshuffle
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8 import (
	generate_gemm_a8w8_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a8w8_blockscale.py

Lines 1 to 34 in 5332c56

	import functools
	import math
	from collections.abc import Callable
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_a8w8_blockscale import (
	gemm_a8w8_blockscale as triton_gemm_a8w8_blockscale,
	)
	from aiter.ops.triton.gemm.basic.gemm_a8w8_blockscale import (
	gemm_a8w8_blockscale_preshuffle as triton_gemm_a8w8_blockscale_preshuffle,
	)
	from aiter.test_common import checkAllclose
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8_blockscale import (
	generate_gemm_a8w8_blockscale_inputs,
	run_torch,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a8w8_per_token_scale.py

Lines 1 to 28 in 5332c56

	import functools
	import math
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_a8w8_per_token_scale import (
	gemm_a8w8_per_token_scale,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8w8_per_token_scale import (
	generate_gemm_a8w8_per_token_scale_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_a8wfp4.py

Lines 1 to 30 in 5332c56

	import functools
	import math
	import sys
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_a8wfp4 import (
	gemm_a8wfp4,
	)
	from aiter.ops.triton.utils.types import get_fp8_dtypes
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_a8wfp4 import (
	generate_gemm_a8wfp4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_afp4wfp4.py

Lines 1 to 31 in 5332c56

	import functools
	import math
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.gemm.basic.gemm_afp4wfp4 import (
	gemm_afp4wfp4 as triton_gemm_afp4wfp4,
	)
	from aiter.ops.triton.gemm.basic.gemm_afp4wfp4 import (
	gemm_afp4wfp4_preshuffle,
	)
	from aiter.ops.triton.utils._triton import arch_info
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)
	from op_tests.triton_tests.gemm.basic.test_gemm_afp4wfp4 import (
	generate_gemm_afp4wfp4_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gemm_afp8wfp8.py

Lines 11 to 40 in 5332c56

	import csv
	import functools
	import math
	import warnings
	from pathlib import Path
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.shuffle import shuffle_weight
	from aiter.ops.triton.gemm.basic.gemm_afp8wfp8 import (
	gemm_afp8wfp8,
	gemm_afp8wfp8_preshuffle,
	)
	from aiter.ops.triton.utils._triton import arch_info
	from op_tests.op_benchmarks.triton.utils.argparse import (
	add_argparse_ff,
	get_do_bench_graph_kwargs,
	get_ff_args,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_model_benchmark_object,
	get_shape_benchmark_object,
	print_vgpr,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_gmm.py

Lines 9 to 64 in 5332c56

	import argparse
	import functools
	import logging
	# PyTorch
	import torch
	# Triton
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	# AITER: Triton kernel wrappers
	from aiter.ops.triton.gemm.grouped.gmm import (
	gmm as triton_gmm,
	)
	from aiter.ops.triton.gemm.grouped.gmm import (
	nptgmm as triton_nptgmm,
	)
	from aiter.ops.triton.gemm.grouped.gmm import (
	ptgmm as triton_ptgmm,
	)
	# AITER: GMM defaults and utility functions
	from aiter.ops.triton.utils.gmm_common import (
	DTYPE,
	DTYPE_STR,
	GROUP_SIZES_DTYPE,
	GROUP_SIZES_DTYPE_STR,
	NUM_GROUP_SIZES,
	RNG_SEED,
	SUPPORTED_DTYPES_STR,
	SUPPORTED_GROUP_SIZES_DTYPES_STR,
	TRANS_LHS,
	TRANS_RHS,
	dtype_from_str,
	gen_gmm_input,
	gen_gmm_output,
	gen_multiple_group_sizes,
	gen_tgmm_bias_grad,
	gen_tgmm_input,
	gen_tgmm_output,
	group_sizes_dtype_from_str,
	str_from_dtype,
	str_from_group_sizes_dtype,
	)
	# AITER: shared argparse helper
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	positive_int,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_layernorm.py

Lines 1 to 27 in 5332c56

	import functools
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.normalization.norm import (
	layer_norm,
	layernorm2d_fwd_with_add,
	layernorm2d_fwd_with_add_dynamicquant,
	layernorm2d_fwd_with_add_smoothquant,
	layernorm2d_fwd_with_dynamicquant,
	layernorm2d_fwd_with_smoothquant,
	)
	from aiter.ops.triton.utils.types import str_to_torch_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	print_vgpr,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_mla.py

Lines 6 to 24 in 5332c56

	import functools
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.attention.mla import mla_decode_fwd, mla_prefill_fwd
	from aiter.ops.triton.utils._triton import arch_info
	from aiter.ops.triton.utils.types import e4m3_dtype
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_caller_name_no_ext
	from op_tests.triton_tests.attention.test_mla import input_helper, shuffle_kv_buffer

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_mla_decode_rope.py

Lines 1 to 29 in 5332c56

	import argparse
	import functools
	import itertools
	import sys
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.attention.mla_decode_rope import (
	decode_attention_fwd_grouped_rope,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_configs,
	print_vgpr,
	)
	from op_tests.triton_tests.attention.test_mla_decode_rope import (
	input_helper,
	ref_preprocess,
	)
	from op_tests.triton_tests.utils.rotary_embedding import DeepseekScalingRotaryEmbedding

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_rmsnorm.py

Lines 1 to 25 in 5332c56

	import functools
	import math
	import torch
	import triton
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.normalization.rmsnorm import rms_norm
	from aiter.ops.triton.quant.fused_mxfp4_quant import fused_rms_mxfp4_quant
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	override_parser_action,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import (
	get_caller_name_no_ext,
	get_model_configs,
	print_vgpr,
	)
	from op_tests.triton_tests.normalization.test_rmsnorm import (
	generate_rmsnorm_inputs,
	)

---

⚠️ [ruff] <I001> _{reported by reviewdog 🐶}
Import block is un-sorted or un-formatted

aiter/op_tests/op_benchmarks/triton/bench_rope.py

Lines 1 to 40 in 5332c56

	import functools
	import torch
	from op_tests.op_benchmarks.triton.utils.rotating_buffer_bench import (
	RotatingBufferBench,
	)
	from aiter.ops.triton.rope.rope import (
	RotateStyle,
	rope_bwd,
	rope_cached_bwd,
	rope_cached_fwd,
	rope_cached_fwd_inplace,
	rope_cached_positions_bwd,
	rope_cached_positions_fwd,
	rope_cached_positions_fwd_inplace,
	rope_cached_positions_offsets_bwd,
	rope_cached_positions_offsets_fwd,
	rope_cached_positions_offsets_fwd_inplace,
	rope_cached_thd_positions_2c_bwd,
	rope_cached_thd_positions_2c_fwd,
	rope_cached_thd_positions_2c_fwd_inplace,
	rope_cached_thd_positions_offsets_2c_bwd,
	# rope_fwd_2d,
	# rope_fwd_2d_inplace,
	rope_cached_thd_positions_offsets_2c_fwd,
	rope_cached_thd_positions_offsets_2c_fwd_inplace,
	rope_fwd,
	rope_fwd_inplace,
	rope_thd_bwd,
	rope_thd_fwd,
	rope_thd_fwd_inplace,
	)
	from op_tests.op_benchmarks.triton.utils.argparse import (
	get_do_bench_graph_kwargs,
	get_parser,
	get_rot_bufs_kwargs,
	)
	from op_tests.op_benchmarks.triton.utils.benchmark_utils import get_model_configs
	from op_tests.triton_tests.rope.test_rope import generate_rope_inputs

[Boss2002n]

The new RotatingBufferBench class is a good cleanup.

Must fix before merge
Two benchmarks no longer run (benchmark_utils.py:378-380): bench_all_reduce.py and bench_all_reduce_rmsnorm_quant.py still import the deleted helpers and fail with ImportError. Fix: move them to RotatingBufferBench.
The default pool is one input set (rotating_buffer_bench.py:82-83): min_pool_len and max_pool_len default to 1, so benchmarks built without the CLI options stop rotating. Fix: default to 4 and 1000.
Settings leak between shapes (rotating_buffer_bench.py:175-184): per-call values stay on the object, so a later shape can get 64 entries (32 GiB) and pass the memory cap. Fix: keep them local.

Nice to fix
Restore del kwargs, v after sizing (rotating_buffer_bench.py:225-232).
Reject unknown keys in from_dict() (rotating_buffer_bench.py:117-126).
Update the --graph-* help, which names the deleted do_bench_aiter_triton (argparse.py:99-122).
Use logger instead of print, shorten the long docstrings, write gfx950 instead of MI355X, and keep main's names (_out_val_keys).

Merge order
Merge this PR first (nine PRs carry its API commit), after a rebase for the conflict with #6727.

[Boss2002n]

LGTM