🚀 Efficient implementations for emerging model architectures
This report summarizes scanner evidence for fla-org/flash-linear-attention, a Python project with 5,842 GitHub stars. SynthScan examined 214,673 lines of code across 815 source files, recording 2813 pattern matches distributed across 21 syntactic categories. The legacy adjusted pattern signal is 17.8 (Elevated pattern signal); it is not an authorship estimate.
Use the findings to ask whether maintainers can explain intent, verify dependencies, simplify unnecessary scaffolding, and focus review effort. Risk Payload v1 adds ownership, supply-chain, over-engineering, and review-effort vectors when available. These are prompts for human verification, not probabilities, security findings, or measured maintenance costs.
Use each vector to guide manual verification of the underlying findings.
Can maintainers explain intent, assumptions, and trade-offs? 1642 finding(s).
Are imports declared and package/API names verified? 705 finding(s).
Are wrappers, guards, and repeated structures necessary? 1715 finding(s).
Which changes need focused human review? 2813 finding(s).
This chart tracks the legacy adjusted pattern score across scan runs. Changes can reflect source changes, scan coverage, or detector/scoring versions; they do not measure AI adoption or prove that a team refactored code. Compare runs only when their scoring versions are compatible. The dashed line tracks the raw pattern-hit count.
Groups findings by configured severity weight. Severity is not a confidence estimate, security rating, or proof that a package is missing. Verify imports against the project's manifests and inspect each finding in its local design context.
This horizontal bar chart shows where configured pattern scores accumulate across directories. Treat the bars as locations to inspect, not as evidence of authorship or code quality. Use the file-level findings to decide whether a directory needs a focused ownership, dependency, complexity, or review check.
The scanner identified 2813 distinct pattern matches across 21 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.
Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/test_dispatch.py | 117 | def test_backend_without_switch(monkeypatch, global_enable, default_enable): | CODE |
| LOW | tests/test_dispatch.py | 144 | def test_registered_backend_switches(monkeypatch, enable_dispatch): | CODE |
| LOW | tests/test_dispatch.py | 241 | def test_dispatch_unknown_owner(monkeypatch, operation): | CODE |
| LOW | tests/test_dispatch.py | 249 | def test_dispatch_import_error(monkeypatch, dependency): | CODE |
| LOW | tests/test_dispatch.py | 404 | def test_load_dependency_error(monkeypatch, lazy_backend): | CODE |
| LOW | tests/test_dispatch.py | 477 | def test_register_import_order(run_python): | CODE |
| LOW | tests/test_utils.py | 16 | def test_assert_close_accepts_close_tensors_and_rejects_mismatch(): | CODE |
| LOW | tests/test_utils.py | 26 | def test_tensor_cache_uses_identity_until_disabled(monkeypatch): | CODE |
| LOW | tests/test_utils.py | 49 | def test_input_guard_makes_tensors_contiguous_except_skipped(): | CODE |
| LOW | tests/test_utils.py | 78 | def test_deprecate_kwarg_renames_old_keyword_and_warns(): | CODE |
| LOW | tests/test_utils.py | 92 | def test_deprecate_kwarg_can_raise_when_both_names_are_set(): | CODE |
| LOW | tests/test_utils.py | 101 | def test_public_utils_are_reexported(): | CODE |
| LOW | tests/test_split_package_release.py | 109 | def test_split_wheels_match_release_contract(tmp_path: Path) -> None: | CODE |
| LOW | tests/test_split_package_release.py | 154 | def test_core_then_extension_install_sequence_without_runtime_deps(tmp_path: Path) -> None: | CODE |
| LOW | tests/test_split_package_release.py | 207 | def test_core_only_import_with_repo_root_on_path(tmp_path: Path) -> None: | STRING |
| LOW | tests/test_split_package_release.py | 247 | def test_split_namespace_across_sys_path_entries_without_runtime_deps(tmp_path: Path) -> None: | STRING |
| LOW | tests/test_split_package_release.py | 273 | def test_full_split_import_contract_when_runtime_dependencies_available(tmp_path: Path) -> None: | STRING |
| LOW | tests/test_public_api.py | 20 | def test_top_level_exports_layers_and_non_config_models(): | CODE |
| LOW | tests/test_public_api.py | 35 | def test_public_imports_preserve_callables(run_python, disabled): | CODE |
| LOW | tests/test_public_api.py | 70 | def test_public_call_defaults(owner, name, defaults): | CODE |
| LOW | tests/test_public_api.py | 85 | def test_legacy_imports_preserve_symbol_identity(legacy, current, names): | CODE |
| LOW | tests/test_public_api.py | 92 | def test_public_function_aliases(): | CODE |
| LOW | tests/test_public_api.py | 136 | def test_legacy_module_pickle_and_state_dict(monkeypatch, legacy, name, kwargs, state_keys): | CODE |
| LOW | tests/test_public_api.py | 163 | def test_normalization_imports_preserve_public_exports(run_python, disabled): | CODE |
| LOW | tests/test_dependency_finder.py | 55 | def test_backend_changes_follow_dispatch_ownership(backend_project, changed, expected): | CODE |
| LOW | tests/test_dependency_finder.py | 67 | def test_backend_changes_follow_shared_kernel_imports(backend_project, backend_file, expected): | CODE |
| LOW | tests/test_dependency_finder.py | 76 | def test_dependencies_follow_reexports_and_aliases(tmp_path): | CODE |
| LOW | tests/test_dependency_finder.py | 96 | def test_dependencies_follow_backend_registration(tmp_path, backend_file): | CODE |
| LOW | tests/test_dependency_finder.py | 148 | def test_backend_changes_follow_repository_ownership(changed, expected, unrelated): | CODE |
| LOW | tests/test_dependency_finder.py | 183 | def test_backend_changes_select_original_test_files(repository_finder, changed, expected): | CODE |
| LOW | tests/layers/test_attn_varlen_pack_layout.py | 37 | def test_attention_varlen_accepts_batched_layout_with_cu_seqlens(B: int, T: int, H: int, D: int): | CODE |
| LOW | tests/layers/test_attn_varlen_pack_layout.py | 126 | def test_layer_prefers_explicit_cu_seqlens_over_attention_mask(layer_cls: type, layer_kwargs: dict) -> None: | CODE |
| LOW | tests/layers/test_cyfa_layer.py | 39 | def test_layer_eval_with_grad_uses_chunk(mode): | CODE |
| LOW | tests/layers/test_cyfa_layer.py | 67 | def test_cached_multi_token_attention_mask(): | CODE |
| LOW | tests/layers/test_gated_deltanet.py | 18 | def test_gated_deltanet_eval_with_grad_uses_chunk(mode: str): | CODE |
| LOW | tests/layers/test_gated_deltanet.py | 52 | def test_gated_deltanet_fused_qkv_conv_matches_fallback(B: int, T: int, D: int, dtype: torch.dtype): | CODE |
| LOW | tests/layers/test_layer_dim_validation.py | 37 | def test_expand_dim_rejects_non_integer(layer_cls): | CODE |
| LOW | tests/layers/test_cu_seqlens_cache_guard.py | 63 | def test_cu_seqlens_rejected_with_non_empty_cache(layer_cls: type, layer_kwargs: dict, requires_flash: bool): | CODE |
| LOW | tests/layers/test_deltaformer_cache.py | 22 | def test_cache_request_rejected(dtype: torch.dtype): | CODE |
| LOW | tests/layers/test_lightnet_decode.py | 21 | def test_masked_single_token_decode(dtype: torch.dtype, use_short_conv: bool, warm_cache: bool): | CODE |
| LOW | tests/layers/test_forgetting_attn_cache.py | 17 | def test_empty_sliding_window_cache_prefill(dtype: torch.dtype): | CODE |
| LOW | tests/layers/test_layer_cache_layer_idx.py | 182 | def test_cache_requires_layer_idx(builder, hidden_states): | CODE |
| LOW | tests/layers/test_layer_cache_layer_idx.py | 218 | def test_layers_without_cache_still_work_with_layer_idx_none(builder, hidden_states): | CODE |
| LOW | tests/layers/test_lightnet_cache.py | 21 | def test_warm_cache_multi_token_continuation( | CODE |
| LOW | tests/layers/test_lightnet_cache.py | 59 | def test_varlen_forward_backward(dtype: torch.dtype, use_short_conv: bool, singleton: bool, layout: str, warm_cache: boo | CODE |
| LOW | tests/layers/test_lightnet_cache.py | 214 | def test_all_padding_forward_backward(use_short_conv: bool, layout: str): | CODE |
| LOW | tests/layers/test_lightnet_cache.py | 252 | def test_warm_cache_ignores_left_padding_content(dtype: torch.dtype, T_continue: int, use_short_conv: bool): | CODE |
| LOW | tests/layers/test_lightnet_cache.py | 278 | def test_warm_cache_all_padding_keeps_state(dtype: torch.dtype): | CODE |
| LOW | tests/layers/test_lightnet_cache.py | 304 | def test_warm_cache_padding_has_zero_gradient(dtype: torch.dtype, T_continue: int): | CODE |
| LOW | tests/layers/test_lightnet_cache.py | 324 | def test_warm_cache_padding_before_first_valid_token(dtype: torch.dtype, T_continue: int): | CODE |
| LOW | tests/layers/test_stickbreaking_attn.py | 88 | def test_rejects_invalid_heads(hidden_size, num_heads, num_kv_heads): | CODE |
| LOW | tests/layers/test_stickbreaking_attn.py | 98 | def test_rejects_unsupported_outputs(kwargs): | CODE |
| LOW | tests/layers/test_stickbreaking_attn.py | 111 | def test_rejects_invalid_layout(): | CODE |
| LOW | tests/layers/test_wall_attn_cache.py | 17 | def test_empty_sliding_window_cache_prefill(dtype: torch.dtype): | CODE |
| LOW | tests/layers/test_wall_attn_cache.py | 46 | def test_cache_with_attention_mask_guard(window_size: int | None): | CODE |
| LOW | tests/layers/test_multiscale_retention_cache.py | 17 | def test_decode_rotary_offset_with_attention_mask(dtype: torch.dtype): | CODE |
| LOW | tests/layers/test_forgetting_attn_dense_decode.py | 17 | def test_dense_decode_without_mask_is_rejected(dtype: torch.dtype): | CODE |
| LOW | tests/layers/test_moba_cache.py | 18 | def test_moba_rejects_cu_seqlens_with_nonempty_cache(): | CODE |
| LOW | tests/layers/test_simple_gla_padding.py | 17 | def test_forward_ignores_padded_gate_content(dtype: torch.dtype): | CODE |
| LOW | tests/layers/test_simple_gla_padding.py | 48 | def test_cached_state_ignores_padded_gate_content(dtype: torch.dtype): | CODE |
| 894 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tests/conftest.py | 33 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/conftest.py | 35 | # ----------------------------------------------------------------------------- | COMMENT |
| MEDIUM | tests/models/test_modeling_deltaformer.py | 22 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_deltaformer.py | 24 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_deltaformer.py | 61 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_deltaformer.py | 63 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_log_linear_mamba2.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_log_linear_mamba2.py | 20 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_yoco.py | 86 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_yoco.py | 88 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_yoco.py | 137 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_yoco.py | 139 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_wall_transformer.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_wall_transformer.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_wall_transformer.py | 54 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_wall_transformer.py | 56 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_gated_deltanet.py | 21 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_gated_deltanet.py | 23 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_gated_deltanet.py | 46 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_gated_deltanet.py | 48 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_gated_deltanet.py | 69 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_gated_deltanet.py | 71 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_path_attn.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_path_attn.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_path_attn.py | 53 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_path_attn.py | 55 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_lightnet.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_lightnet.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_lightnet.py | 56 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_lightnet.py | 58 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_gsa.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_gsa.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_gsa.py | 55 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_gsa.py | 57 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_hgrn.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_hgrn.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_hgrn.py | 55 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_hgrn.py | 57 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_mom.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_mom.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_mom.py | 56 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_mom.py | 58 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_kda.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_kda.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_kda.py | 39 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_precond_kda.py | 41 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_kda.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_kda.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_kda.py | 54 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_kda.py | 56 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_deltanet.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_deltanet.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_deltanet.py | 55 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_deltanet.py | 57 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_forgetting_transformer.py | 16 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_forgetting_transformer.py | 18 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_forgetting_transformer.py | 55 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_forgetting_transformer.py | 57 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_gated_deltaproduct.py | 19 | # =================================================================================== | COMMENT |
| MEDIUM | tests/models/test_modeling_gated_deltaproduct.py | 21 | # =================================================================================== | COMMENT |
| 179 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | evals/harness.py | 8 | CODE | |
| LOW | evals/harness.py | 10 | CODE | |
| LOW | evals/ppl.py | 25 | CODE | |
| LOW | tests/layers/test_attn_varlen_pack_layout.py | 30 | CODE | |
| LOW | tests/layers/test_cu_seqlens_cache_guard.py | 20 | CODE | |
| LOW | tests/models/test_modeling_yoco.py | 16 | CODE | |
| LOW | …s/skills/fla-ascend-performance/scripts/profile_npu.py | 35 | CODE | |
| LOW | …ills/fla-ascend-performance/scripts/analyze_profile.py | 17 | CODE | |
| LOW | legacy/training/run.py | 14 | CODE | |
| LOW | legacy/training/preprocess.py | 8 | CODE | |
| LOW | legacy/training/flame/parser.py | 8 | CODE | |
| LOW | legacy/training/flame/data.py | 8 | CODE | |
| LOW | utils/convert_from_llama.py | 37 | CODE | |
| LOW | utils/convert_from_rwkv7.py | 17 | CODE | |
| LOW | utils/convert_from_rwkv6.py | 35 | CODE | |
| LOW | benchmarks/visualize.py | 27 | CODE | |
| LOW | benchmarks/benchmark_generation.py | 15 | CODE | |
| LOW | benchmarks/ops/run.py | 139 | CODE | |
| LOW | benchmarks/ops/registry.py | 14 | CODE | |
| LOW | benchmarks/ops/graph_benchmark.py | 24 | CODE | |
| LOW | benchmarks/ops/verify.py | 52 | CODE | |
| LOW | benchmarks/ops/graph_registry.py | 10 | CODE | |
| LOW | benchmarks/modules/runner.py | 10 | CODE | |
| LOW | scripts/utils/__init__.py | 8 | CODE | |
| LOW | scripts/utils/__init__.py | 9 | CODE | |
| LOW | scripts/utils/__init__.py | 9 | CODE | |
| LOW | fla/backends.py | 10 | CODE | |
| LOW | fla/layers/precond_kda.py | 8 | CODE | |
| LOW | fla/layers/hgrn.py | 10 | CODE | |
| LOW | fla/layers/raven.py | 11 | CODE | |
| LOW | fla/layers/momentum_deltanet.py | 8 | CODE | |
| LOW | fla/layers/precond_gated_deltanet.py | 8 | CODE | |
| LOW | fla/layers/hgrn2.py | 10 | CODE | |
| LOW | fla/layers/deltaformer.py | 8 | CODE | |
| LOW | fla/layers/mom.py | 8 | CODE | |
| LOW | fla/layers/delta_net.py | 8 | CODE | |
| LOW | fla/layers/mamba.py | 8 | CODE | |
| LOW | fla/layers/kda.py | 8 | CODE | |
| LOW | fla/layers/bitattn.py | 8 | CODE | |
| LOW | fla/layers/mamba2.py | 8 | CODE | |
| LOW | fla/layers/rwkv7.py | 8 | CODE | |
| LOW | fla/layers/stickbreaking_attn.py | 8 | CODE | |
| LOW | fla/layers/yoco.py | 8 | CODE | |
| LOW | fla/layers/gsa.py | 8 | CODE | |
| LOW | fla/layers/parallax.py | 11 | CODE | |
| LOW | fla/layers/mamba3.py | 8 | CODE | |
| LOW | fla/layers/rwkv6.py | 10 | CODE | |
| LOW | fla/layers/__init__.py | 8 | CODE | |
| LOW | fla/layers/__init__.py | 9 | CODE | |
| LOW | fla/layers/__init__.py | 10 | CODE | |
| LOW | fla/layers/__init__.py | 11 | CODE | |
| LOW | fla/layers/__init__.py | 12 | CODE | |
| LOW | fla/layers/__init__.py | 13 | CODE | |
| LOW | fla/layers/__init__.py | 14 | CODE | |
| LOW | fla/layers/__init__.py | 15 | CODE | |
| LOW | fla/layers/__init__.py | 16 | CODE | |
| LOW | fla/layers/__init__.py | 17 | CODE | |
| LOW | fla/layers/__init__.py | 18 | CODE | |
| LOW | fla/layers/__init__.py | 19 | CODE | |
| LOW | fla/layers/__init__.py | 20 | CODE | |
| 630 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/models/test_modeling_cat.py | 19 | CODE | |
| LOW | tests/models/test_modeling_mamba2.py | 142 | CODE | |
| LOW | tests/context_parallel/test_cp_kda.py | 420 | CODE | |
| LOW | …s/skills/fla-ascend-performance/scripts/profile_npu.py | 74 | CODE | |
| LOW | legacy/training/preprocess.py | 69 | CODE | |
| LOW | benchmarks/benchmark_training_throughput.py | 88 | CODE | |
| LOW | benchmarks/cp/test_gdn_with_cp.py | 257 | CODE | |
| LOW | benchmarks/cp/test_gdn_with_cp.py | 269 | CODE | |
| LOW | benchmarks/cp/test_gdn_with_cp.py | 351 | CODE | |
| LOW | benchmarks/cp/test_gdn_with_cp.py | 503 | CODE | |
| LOW | fla/layers/precond_kda.py | 92 | CODE | |
| LOW | fla/layers/hgrn.py | 31 | CODE | |
| LOW | fla/layers/raven.py | 44 | CODE | |
| LOW | fla/layers/momentum_deltanet.py | 37 | CODE | |
| LOW | fla/layers/precond_gated_deltanet.py | 104 | CODE | |
| LOW | fla/layers/hgrn2.py | 33 | CODE | |
| LOW | fla/layers/deltaformer.py | 62 | CODE | |
| LOW | fla/layers/mom.py | 295 | CODE | |
| LOW | fla/layers/delta_net.py | 75 | CODE | |
| LOW | fla/layers/mamba.py | 54 | CODE | |
| LOW | fla/layers/kda.py | 83 | CODE | |
| LOW | fla/layers/bitattn.py | 40 | CODE | |
| LOW | fla/layers/mamba2.py | 113 | CODE | |
| LOW | fla/layers/rwkv7.py | 34 | CODE | |
| LOW | fla/layers/stickbreaking_attn.py | 46 | CODE | |
| LOW | fla/layers/yoco.py | 95 | CODE | |
| LOW | fla/layers/yoco.py | 238 | CODE | |
| LOW | fla/layers/yoco.py | 330 | CODE | |
| LOW | fla/layers/gsa.py | 33 | CODE | |
| LOW | fla/layers/parallax.py | 64 | CODE | |
| LOW | fla/layers/mamba3.py | 64 | CODE | |
| LOW | fla/layers/rwkv6.py | 32 | CODE | |
| LOW | fla/layers/wall_attn.py | 69 | CODE | |
| LOW | fla/layers/rebased.py | 25 | CODE | |
| LOW | fla/layers/path_attn.py | 31 | CODE | |
| LOW | fla/layers/simple_gla.py | 69 | CODE | |
| LOW | fla/layers/gla.py | 76 | CODE | |
| LOW | fla/layers/cyfa.py | 75 | CODE | |
| LOW | fla/layers/comba.py | 82 | CODE | |
| LOW | fla/layers/log_linear_mamba2.py | 43 | CODE | |
| LOW | fla/layers/log_linear_mamba2.py | 283 | CODE | |
| LOW | fla/layers/log_linear_mamba2.py | 131 | CODE | |
| LOW | fla/layers/log_linear_mamba2.py | 294 | CODE | |
| LOW | fla/layers/rodimus.py | 64 | CODE | |
| LOW | fla/layers/rodimus.py | 242 | CODE | |
| LOW | fla/layers/forgetting_attn.py | 32 | CODE | |
| LOW | fla/layers/gated_deltaproduct.py | 35 | CODE | |
| LOW | fla/layers/moba.py | 92 | CODE | |
| LOW | fla/layers/lightnet.py | 32 | CODE | |
| LOW | fla/layers/linear_attn.py | 29 | CODE | |
| LOW | fla/layers/mesa_net.py | 59 | CODE | |
| LOW | fla/layers/mla.py | 55 | CODE | |
| LOW | fla/layers/multiscale_retention.py | 71 | CODE | |
| LOW | fla/layers/nsa.py | 30 | CODE | |
| LOW | fla/layers/attn.py | 39 | CODE | |
| LOW | fla/layers/gated_deltanet.py | 88 | CODE | |
| LOW | fla/layers/abc.py | 29 | CODE | |
| LOW | fla/layers/gdn2.py | 88 | CODE | |
| LOW | fla/models/utils.py | 227 | CODE | |
| LOW | fla/models/utils.py | 399 | CODE | |
| 235 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | tests/models/test_modeling_log_linear_mamba2.py | 0 | test the forward and backward pass of the mamba3 model by manually instantiating the configuration and the model. | STRING |
| HIGH | tests/models/test_modeling_mamba2.py | 0 | test the forward and backward pass of the mamba3 model by manually instantiating the configuration and the model. | STRING |
| HIGH | tests/models/test_modeling_mamba3.py | 0 | test the forward and backward pass of the mamba3 model by manually instantiating the configuration and the model. | STRING |
| HIGH | tests/context_parallel/test_cp_gdp.py | 0 | initialize distributed environment for a single process. | STRING |
| HIGH | tests/context_parallel/test_cp_conv.py | 0 | initialize distributed environment for a single process. | STRING |
| HIGH | tests/context_parallel/test_cp_token_shift.py | 0 | initialize distributed environment for a single process. | STRING |
| HIGH | tests/context_parallel/test_cp_dplr.py | 0 | initialize distributed environment for a single process. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn2.py | 0 | initialize distributed environment for a single process. | STRING |
| HIGH | tests/context_parallel/test_cp_kda.py | 0 | initialize distributed environment for a single process. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn.py | 0 | initialize distributed environment for a single process. | STRING |
| HIGH | tests/context_parallel/test_cp_gdp.py | 0 | worker function for cp gdn test. runs in a spawned process with the given rank. | STRING |
| HIGH | tests/context_parallel/test_cp_conv.py | 0 | worker function for cp gdn test. runs in a spawned process with the given rank. | STRING |
| HIGH | tests/context_parallel/test_cp_dplr.py | 0 | worker function for cp gdn test. runs in a spawned process with the given rank. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn2.py | 0 | worker function for cp gdn test. runs in a spawned process with the given rank. | STRING |
| HIGH | tests/context_parallel/test_cp_kda.py | 0 | worker function for cp gdn test. runs in a spawned process with the given rank. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn.py | 0 | worker function for cp gdn test. runs in a spawned process with the given rank. | STRING |
| HIGH | tests/context_parallel/test_cp_conv.py | 0 | run cp test using torch.multiprocessing.spawn. this allows running the test directly with pytest. | STRING |
| HIGH | tests/context_parallel/test_cp_dplr.py | 0 | run cp test using torch.multiprocessing.spawn. this allows running the test directly with pytest. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn2.py | 0 | run cp test using torch.multiprocessing.spawn. this allows running the test directly with pytest. | STRING |
| HIGH | tests/context_parallel/test_cp_kda.py | 0 | run cp test using torch.multiprocessing.spawn. this allows running the test directly with pytest. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn.py | 0 | run cp test using torch.multiprocessing.spawn. this allows running the test directly with pytest. | STRING |
| HIGH | tests/context_parallel/test_cp_token_shift.py | 0 | cp2: sequence boundaries aligned with rank boundaries. | STRING |
| HIGH | tests/context_parallel/test_cp_dplr.py | 0 | cp2: sequence boundaries aligned with rank boundaries. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn2.py | 0 | cp2: sequence boundaries aligned with rank boundaries. | STRING |
| HIGH | tests/context_parallel/test_cp_kda.py | 0 | cp2: sequence boundaries aligned with rank boundaries. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn.py | 0 | cp2: sequence boundaries aligned with rank boundaries. | STRING |
| HIGH | tests/context_parallel/test_cp_dplr.py | 0 | cp4: complex sequence distribution, first sequence spans 3 ranks. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn2.py | 0 | cp4: complex sequence distribution, first sequence spans 3 ranks. | STRING |
| HIGH | tests/context_parallel/test_cp_kda.py | 0 | cp4: complex sequence distribution, first sequence spans 3 ranks. | STRING |
| HIGH | tests/context_parallel/test_cp_gdn.py | 0 | cp4: complex sequence distribution, first sequence spans 3 ranks. | STRING |
| HIGH | fla/layers/mamba.py | 0 | compute ∆, a, b, c, and d the state space parameters and compute the `contextualized_states`. a, d are input independent | STRING |
| HIGH | fla/layers/mamba2.py | 0 | compute ∆, a, b, c, and d the state space parameters and compute the `contextualized_states`. a, d are input independent | STRING |
| HIGH | fla/layers/log_linear_mamba2.py | 0 | compute ∆, a, b, c, and d the state space parameters and compute the `contextualized_states`. a, d are input independent | STRING |
| HIGH | …models/log_linear_mamba2/modeling_log_linear_mamba2.py | 0 | an abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. | STRING |
| HIGH | fla/models/mamba/modeling_mamba.py | 0 | an abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. | STRING |
| HIGH | fla/models/mamba2/modeling_mamba2.py | 0 | an abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. | STRING |
| HIGH | fla/models/samba/modeling_samba.py | 0 | an abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. | STRING |
| HIGH | fla/ops/retention/chunk.py | 0 | args: q (torch.tensor): queries of shape `[b, t, h, k]`. k (torch.tensor): keys of shape `[b, t, h, k]`. v (torch.tensor | STRING |
| HIGH | fla/ops/retention/fused_chunk.py | 0 | args: q (torch.tensor): queries of shape `[b, t, h, k]`. k (torch.tensor): keys of shape `[b, t, h, k]`. v (torch.tensor | STRING |
| HIGH | fla/ops/generalized_delta_rule/iplr/chunk.py | 0 | args: q (torch.tensor): queries of shape `[b, t, h, k]`. k (torch.tensor): keys of shape `[b, t, h, k]`. v (torch.tensor | STRING |
| HIGH | fla/ops/linear_attn/chunk.py | 0 | args: q (torch.tensor): queries of shape `[b, t, h, k]`. k (torch.tensor): keys of shape `[b, t, h, k]`. v (torch.tensor | STRING |
| HIGH | fla/ops/linear_attn/fused_chunk.py | 0 | args: q (torch.tensor): queries of shape `[b, t, h, k]`. k (torch.tensor): keys of shape `[b, t, h, k]`. v (torch.tensor | STRING |
| HIGH | fla/ops/linear_attn/fused_recurrent.py | 0 | args: q (torch.tensor): queries of shape `[b, t, h, k]`. k (torch.tensor): keys of shape `[b, t, h, k]`. v (torch.tensor | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/conftest.py | 58 | CODE | |
| LOW | tests/models/test_modeling_utils.py | 55 | CODE | |
| LOW | tests/ops/test_nsa.py | 38 | CODE | |
| LOW | tests/ops/test_gdn2.py | 566 | CODE | |
| LOW | tests/ops/test_dplr_tilelang.py | 205 | CODE | |
| LOW | tests/ops/utils/test_csr.py | 47 | CODE | |
| LOW | tests/context_parallel/test_cp_gdp.py | 52 | CODE | |
| LOW | tests/context_parallel/test_cp_dplr.py | 75 | CODE | |
| LOW | tests/context_parallel/test_cp_gdn2.py | 107 | CODE | |
| LOW | tests/context_parallel/test_cp_kda.py | 103 | CODE | |
| LOW | tests/context_parallel/test_cp_gdn.py | 100 | CODE | |
| LOW | tests/modules/test_conv.py | 1241 | CODE | |
| LOW | legacy/training/flame/data.py | 186 | CODE | |
| LOW | benchmarks/benchmark_training_throughput.py | 88 | CODE | |
| LOW | benchmarks/visualize.py | 150 | CODE | |
| LOW | benchmarks/ops/run.py | 280 | CODE | |
| LOW | benchmarks/ops/registry.py | 214 | CODE | |
| LOW | benchmarks/modules/runner.py | 41 | CODE | |
| LOW | benchmarks/modules/benchmark_cross_entropy.py | 37 | CODE | |
| LOW | benchmarks/modules/benchmark_conv.py | 40 | CODE | |
| LOW | benchmarks/modules/benchmark_layernorm.py | 51 | CODE | |
| LOW | benchmarks/modules/benchmark_activations.py | 71 | CODE | |
| LOW | scripts/find_dependent_tests.py | 39 | CODE | |
| LOW | scripts/find_dependent_tests.py | 67 | CODE | |
| LOW | scripts/find_dependent_tests.py | 136 | CODE | |
| LOW | scripts/find_dependent_tests.py | 266 | CODE | |
| LOW | scripts/run_benchmark_compare.py | 79 | CODE | |
| LOW | scripts/run_benchmark_compare.py | 209 | CODE | |
| LOW | scripts/run_benchmark_compare.py | 310 | CODE | |
| LOW | scripts/utils/autotune_export.py | 241 | CODE | |
| LOW | scripts/utils/autotune_export.py | 192 | CODE | |
| LOW | fla/layers/momentum_deltanet.py | 186 | CODE | |
| LOW | fla/layers/delta_net.py | 170 | CODE | |
| LOW | fla/layers/mamba2.py | 280 | CODE | |
| LOW | fla/layers/rwkv6.py | 217 | CODE | |
| LOW | fla/layers/wall_attn.py | 133 | CODE | |
| LOW | fla/layers/log_linear_mamba2.py | 43 | CODE | |
| LOW | fla/layers/log_linear_mamba2.py | 131 | CODE | |
| LOW | fla/layers/log_linear_mamba2.py | 453 | CODE | |
| LOW | fla/layers/linear_attn.py | 29 | CODE | |
| LOW | fla/layers/multiscale_retention.py | 175 | CODE | |
| LOW | fla/utils/ascend_ub_manager.py | 72 | CODE | |
| LOW | fla/models/utils.py | 42 | CODE | |
| LOW | fla/models/utils.py | 227 | CODE | |
| LOW | fla/models/utils.py | 468 | CODE | |
| LOW | fla/models/utils.py | 498 | CODE | |
| LOW | fla/models/kda/modeling_kda.py | 198 | CODE | |
| LOW | fla/models/kda/modeling_kda.py | 406 | CODE | |
| LOW | fla/models/cat/modeling_cat.py | 711 | CODE | |
| LOW | fla/models/abc/modeling_abc.py | 188 | CODE | |
| LOW | fla/models/abc/modeling_abc.py | 377 | CODE | |
| LOW | fla/models/delta_net/modeling_delta_net.py | 188 | CODE | |
| LOW | fla/models/delta_net/modeling_delta_net.py | 377 | CODE | |
| LOW | …dels/gated_deltaproduct/modeling_gated_deltaproduct.py | 187 | CODE | |
| LOW | …dels/gated_deltaproduct/modeling_gated_deltaproduct.py | 379 | CODE | |
| LOW | fla/models/gsa/modeling_gsa.py | 191 | CODE | |
| LOW | fla/models/gsa/modeling_gsa.py | 381 | CODE | |
| LOW | fla/models/moba/modeling_moba.py | 172 | CODE | |
| LOW | fla/models/moba/modeling_moba.py | 344 | CODE | |
| LOW | fla/models/cyfa/modeling_cyfa.py | 117 | CODE | |
| 139 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| CRITICAL | utils/convert_from_llama.py | 170 | model.model.norm.weight.data.copy_(llama.model.norm.weight) | CODE |
| CRITICAL | utils/convert_from_llama.py | 174 | model.model.norm.bias.data.copy_(llama.model.norm.bias) | CODE |
| CRITICAL | utils/convert_from_rwkv6.py | 65 | model.model.embeddings.weight.data.copy_(rwkv6.rwkv.embeddings.weight) | CODE |
| CRITICAL | utils/convert_from_rwkv6.py | 114 | model.model.layers[i].attn.r_proj.linear.weight.data.copy_(rwkv6.rwkv.blocks[i].attention.receptance.weight) | CODE |
| CRITICAL⚡ | utils/convert_from_rwkv6.py | 131 | model.model.layers[i].attn.k_proj.linear.weight.data.copy_(rwkv6.rwkv.blocks[i].attention.key.weight) | CODE |
| CRITICAL⚡ | utils/convert_from_rwkv6.py | 135 | model.model.layers[i].attn.v_proj.linear.weight.data.copy_(rwkv6.rwkv.blocks[i].attention.value.weight) | CODE |
| CRITICAL⚡ | utils/convert_from_rwkv6.py | 139 | model.model.layers[i].attn.g_proj.linear.weight.data.copy_(rwkv6.rwkv.blocks[i].attention.gate.weight) | CODE |
| CRITICAL | utils/convert_from_rwkv6.py | 172 | model.model.layers[i].ffn.key.linear.weight.data.copy_(rwkv6.rwkv.blocks[i].feed_forward.key.weight) | CODE |
| CRITICAL⚡ | utils/convert_from_rwkv6.py | 186 | model.model.layers[i].ffn.receptance.linear.weight.data.copy_(rwkv6.rwkv.blocks[i].feed_forward.receptance.weigh | CODE |
| CRITICAL⚡ | utils/convert_from_rwkv6.py | 196 | model.model.norm.weight.data.copy_(rwkv6.rwkv.ln_out.weight) | CODE |
| CRITICAL⚡ | utils/convert_from_rwkv6.py | 200 | model.model.norm.bias.data.copy_(rwkv6.rwkv.ln_out.bias) | CODE |
| CRITICAL⚡ | fla/utils/hardware.py | 33 | return triton.runtime.driver.active.utils.get_device_properties(tensor_idx)['multiprocessor_count'] | CODE |
| CRITICAL⚡ | fla/utils/hardware.py | 38 | props = triton.runtime.driver.active.utils.get_device_properties(tensor_idx) | CODE |
| CRITICAL | fla/utils/hardware.py | 153 | triton.runtime.driver.active.utils.get_device_properties(i)['max_shared_mem'] | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | legacy/training/preprocess.py | 29 | Tokenize the input text and split into chunks of specified context length. Args: examples: | STRING |
| HIGH | fla/utils/ascend_ub_manager.py | 258 | Compute UB-safe tiled shapes, padding non-tiling dimensions to powers of two. Args: safety_margin (float, O | STRING |
| HIGH | fla/utils/decorators.py | 195 | Decorator to notify users about deprecated keyword arguments, replacing them with a new name if specified. Thi | STRING |
| HIGH | fla/ops/kda/chunk.py | 241 | Args: q (torch.Tensor): queries of shape ``[B, T, H, K]``. k (torch.Tensor): | STRING |
| HIGH | fla/ops/kda/fused_recurrent.py | 359 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/delta_rule/chunk.py | 240 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/delta_rule/fused_recurrent.py | 454 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/gated_delta_rule/chunk.py | 416 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/gated_delta_rule/fused_recurrent.py | 330 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/gsa/chunk.py | 1143 | Args: q (torch.Tensor): queries of shape `[B, T, HQ, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/gsa/fused_recurrent.py | 450 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/moba/parallel.py | 305 | Flash-attn based MoBA implementation. MoBA (Mixture of Block Attention, https://arxiv.org/abs/2502.13189) makes | STRING |
| HIGH | fla/ops/hgrn/fused_recurrent.py | 269 | Args: x (torch.Tensor): inputs of shape `[B, T, D]. g (torch.Tensor): Forge | STRING |
| HIGH | fla/ops/gated_delta_product/chunk.py | 295 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/precond_kda/chunk.py | 594 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/precond_kda/fused_recurrent.py | 442 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/precond_gated_delta_rule/chunk.py | 430 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/precond_gated_delta_rule/fused_recurrent.py | 369 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/simple_gla/chunk.py | 252 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/simple_gla/fused_chunk.py | 26 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/simple_gla/fused_recurrent.py | 31 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/gla/chunk.py | 1466 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/gla/fused_recurrent.py | 29 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/mesa_net/chunk.py | 322 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]` k (torch.Tensor): k | STRING |
| HIGH | fla/ops/gdn2/chunk.py | 226 | Chunkwise forward for Gated DeltaNet 2 (GDN-2). The matrix-state recurrence on ``S in R^{K x V}`` is | STRING |
| HIGH | fla/ops/comba/chunk.py | 308 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/comba/fused_recurrent.py | 238 | Args: q (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/rwkv6/chunk.py | 1291 | Args: r (torch.Tensor): queries of shape `[B, T, H, K]`. k (torch.Tensor): | STRING |
| HIGH | fla/ops/rwkv6/fused_recurrent.py | 587 | Args: r (torch.Tensor): reception of shape `[B, T, H, K]`. Alias: q, query in linea | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | benchmarks/ops/run.py | 168 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | benchmarks/ops/registry.py | 24 | logger = logging.getLogger(__name__) | CODE |
| LOW | benchmarks/ops/graph_benchmark.py | 45 | logger = logging.getLogger(__name__) | CODE |
| LOW | scripts/utils/__init__.py | 11 | __all__ = [ | CODE |
| LOW | scripts/utils/autotune_export.py | 137 | def update_default_config(self) -> None: | CODE |
| LOW | fla/backends.py | 24 | logger = logging.getLogger(__name__) | CODE |
| LOW | fla/backends.py | 266 | __all__ = ['BackendRegistry', 'BaseBackend', 'dispatch', 'register'] | CODE |
| LOW | fla/layers/__init__.py | 51 | __all__ = [ | CODE |
| LOW | fla/utils/env.py | 23 | logger = logging.getLogger(__name__) | CODE |
| LOW⚡ | fla/utils/hardware.py | 23 | logger = logging.getLogger(__name__) | CODE |
| LOW | fla/utils/testing.py | 15 | logger = logging.getLogger(__name__) | CODE |
| LOW | fla/models/__init__.py | 65 | __all__ = [ | CODE |
| LOW | fla/models/kda/__init__.py | 17 | __all__ = ['KDAConfig', 'KDAForCausalLM', 'KDAModel'] | CODE |
| LOW | fla/models/cat/__init__.py | 20 | __all__ = ['CATConfig', 'CATForCausalLM', 'CATModel'] | CODE |
| LOW | fla/models/abc/__init__.py | 18 | __all__ = ['ABCConfig', 'ABCForCausalLM', 'ABCModel'] | CODE |
| LOW | fla/models/delta_net/__init__.py | 17 | __all__ = ['DeltaNetConfig', 'DeltaNetForCausalLM', 'DeltaNetModel'] | CODE |
| LOW | fla/models/gated_deltaproduct/__init__.py | 17 | __all__ = [ | CODE |
| LOW | fla/models/gsa/__init__.py | 18 | __all__ = ['GSAConfig', 'GSAForCausalLM', 'GSAModel'] | CODE |
| LOW | fla/models/moba/__init__.py | 18 | __all__ = ['MoBAConfig', 'MoBAForCausalLM', 'MoBAModel'] | CODE |
| LOW | fla/models/cyfa/__init__.py | 23 | __all__ = [ | CODE |
| LOW | fla/models/log_linear_mamba2/__init__.py | 18 | __all__ = ['LogLinearMamba2Config', 'LogLinearMamba2ForCausalLM', 'LogLinearMamba2Model'] | CODE |
| LOW | fla/models/deltaformer/__init__.py | 17 | __all__ = ['DeltaFormerConfig', 'DeltaFormerForCausalLM', 'DeltaFormerModel'] | CODE |
| LOW | fla/models/hgrn/__init__.py | 18 | __all__ = ['HGRNConfig', 'HGRNForCausalLM', 'HGRNModel'] | CODE |
| LOW | fla/models/transformer/__init__.py | 18 | __all__ = ['TransformerConfig', 'TransformerForCausalLM', 'TransformerModel'] | CODE |
| LOW | fla/models/yoco/__init__.py | 17 | __all__ = ['YOCOConfig', 'YOCOForCausalLM', 'YOCOModel'] | CODE |
| LOW | fla/models/mamba/__init__.py | 18 | __all__ = ['MambaConfig', 'MambaForCausalLM', 'MambaModel'] | CODE |
| LOW | fla/models/precond_kda/__init__.py | 17 | __all__ = ['PrecondKDAConfig', 'PrecondKDAForCausalLM', 'PrecondKDAModel'] | CODE |
| LOW | fla/models/nsa/__init__.py | 18 | __all__ = ['NSAConfig', 'NSAForCausalLM', 'NSAModel'] | CODE |
| LOW | fla/models/lightnet/__init__.py | 18 | __all__ = ['LightNetConfig', 'LightNetForCausalLM', 'LightNetModel'] | CODE |
| LOW | fla/models/path_attn/__init__.py | 18 | __all__ = ['PaTHAttentionConfig', 'PaTHAttentionForCausalLM', 'PaTHAttentionModel'] | CODE |
| LOW | fla/models/parallax/__init__.py | 24 | __all__ = ['ParallaxConfig', 'ParallaxForCausalLM', 'ParallaxModel'] | CODE |
| LOW | fla/models/gla/__init__.py | 18 | __all__ = ['GLAConfig', 'GLAForCausalLM', 'GLAModel'] | CODE |
| LOW | fla/models/mla/__init__.py | 18 | __all__ = ['MLAConfig', 'MLAForCausalLM', 'MLAModel'] | CODE |
| LOW | fla/models/raven/__init__.py | 18 | __all__ = ['RavenConfig', 'RavenForCausalLM', 'RavenModel'] | CODE |
| LOW | fla/models/mamba2/__init__.py | 18 | __all__ = ['Mamba2Config', 'Mamba2ForCausalLM', 'Mamba2Model'] | CODE |
| LOW | fla/models/mamba3/__init__.py | 18 | __all__ = ['Mamba3Config', 'Mamba3ForCausalLM', 'Mamba3Model'] | CODE |
| LOW | fla/models/samba/__init__.py | 18 | __all__ = ['SambaConfig', 'SambaForCausalLM', 'SambaModel'] | CODE |
| LOW | fla/models/mesa_net/__init__.py | 17 | __all__ = ['MesaNetConfig', 'MesaNetForCausalLM', 'MesaNetModel'] | CODE |
| LOW | fla/models/linear_attn/__init__.py | 17 | __all__ = ['LinearAttentionConfig', 'LinearAttentionForCausalLM', 'LinearAttentionModel'] | CODE |
| LOW | fla/models/wall_transformer/__init__.py | 23 | __all__ = ['WallTransformerConfig', 'WallTransformerForCausalLM', 'WallTransformerModel'] | CODE |
| LOW | fla/models/retnet/__init__.py | 18 | __all__ = ['RetNetConfig', 'RetNetForCausalLM', 'RetNetModel'] | CODE |
| LOW | fla/models/gated_deltanet/__init__.py | 17 | __all__ = ['GatedDeltaNetConfig', 'GatedDeltaNetForCausalLM', 'GatedDeltaNetModel'] | CODE |
| LOW | fla/models/forgetting_transformer/__init__.py | 21 | __all__ = ['ForgettingTransformerConfig', 'ForgettingTransformerForCausalLM', 'ForgettingTransformerModel'] | CODE |
| LOW | fla/models/hgrn2/__init__.py | 18 | __all__ = ['HGRN2Config', 'HGRN2ForCausalLM', 'HGRN2Model'] | CODE |
| LOW | fla/models/stickbreaking_attn/__init__.py | 20 | __all__ = ['StickBreakingAttentionConfig', 'StickBreakingAttentionForCausalLM', 'StickBreakingAttentionModel'] | CODE |
| LOW | fla/models/rodimus/__init__.py | 18 | __all__ = ['RodimusConfig', 'RodimusForCausalLM', 'RodimusModel'] | CODE |
| LOW | fla/models/comba/__init__.py | 17 | __all__ = ['CombaConfig', 'CombaForCausalLM', 'CombaModel'] | CODE |
| LOW | fla/models/rwkv7/__init__.py | 18 | __all__ = ['RWKV7Config', 'RWKV7ForCausalLM', 'RWKV7Model'] | CODE |
| LOW | fla/models/bitnet/__init__.py | 18 | __all__ = ['BitNetConfig', 'BitNetForCausalLM', 'BitNetModel'] | CODE |
| LOW | fla/models/mom/__init__.py | 17 | __all__ = ['MomConfig', 'MomForCausalLM', 'MomModel'] | CODE |
| LOW | fla/models/precond_gated_deltanet/__init__.py | 20 | __all__ = ['PrecondGatedDeltaNetConfig', 'PrecondGatedDeltaNetForCausalLM', 'PrecondGatedDeltaNetModel'] | CODE |
| LOW | fla/models/rwkv6/__init__.py | 18 | __all__ = ['RWKV6Config', 'RWKV6ForCausalLM', 'RWKV6Model'] | CODE |
| LOW | fla/ops/__init__.py | 47 | __all__ = [ | CODE |
| LOW | fla/ops/kda/__init__.py | 11 | __all__ = [ | CODE |
| LOW | fla/ops/kda/backends/__init__.py | 13 | __all__ = ['KDATileLangBackend', 'TritonAscendKDABackend'] | CODE |
| LOW | fla/ops/abc/__init__.py | 10 | __all__ = [ | CODE |
| LOW | fla/ops/delta_rule/__init__.py | 12 | __all__ = [ | CODE |
| LOW | fla/ops/gated_delta_rule/__init__.py | 12 | __all__ = [ | CODE |
| LOW | fla/ops/gated_delta_rule/backends/__init__.py | 13 | __all__ = ['FlashQLABackend', 'TritonAscendGDNBackend'] | CODE |
| LOW | fla/ops/gsa/__init__.py | 11 | __all__ = [ | CODE |
| 80 more matches not shown… | ||||
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/context_parallel/test_cp_gdp.py | 80 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to all) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdp.py | 114 | # Step 2: Reference Run (recurrent, varlen, no CP) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdp.py | 148 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_gdp.py | 187 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_conv.py | 119 | # Step 1: Prepare Global Data | COMMENT |
| LOW | tests/context_parallel/test_cp_conv.py | 137 | # Step 2: Reference Run | COMMENT |
| LOW | tests/context_parallel/test_cp_conv.py | 161 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_conv.py | 205 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW⚡ | tests/context_parallel/test_cp_token_shift.py | 94 | # Step 1: Prepare Global Data | COMMENT |
| LOW⚡ | tests/context_parallel/test_cp_token_shift.py | 104 | # Step 2: Reference Run | COMMENT |
| LOW⚡ | tests/context_parallel/test_cp_token_shift.py | 114 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_token_shift.py | 141 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_dplr.py | 107 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to all) | COMMENT |
| LOW | tests/context_parallel/test_cp_dplr.py | 147 | # Step 2: Reference Run using dplr_recurrence (ground truth) | COMMENT |
| LOW | tests/context_parallel/test_cp_dplr.py | 224 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_dplr.py | 283 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn2.py | 141 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to other ranks) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn2.py | 214 | # Step 2: Reference Run using naive_recurrent_gdn2 (ground truth) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn2.py | 306 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn2.py | 352 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_kda.py | 141 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to other ranks) | COMMENT |
| LOW | tests/context_parallel/test_cp_kda.py | 210 | # Step 2: Reference Run using naive_recurrent_kda (ground truth) | COMMENT |
| LOW | tests/context_parallel/test_cp_kda.py | 308 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_kda.py | 367 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn.py | 130 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to all) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn.py | 163 | # Step 2: Reference Run (single GPU, varlen, no CP) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn.py | 193 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn.py | 229 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | scripts/run_benchmark_compare.py | 371 | # Step 1: HEAD | COMMENT |
| LOW | scripts/run_benchmark_compare.py | 394 | # Step 2: BASE | COMMENT |
| LOW⚡ | scripts/run_benchmark_compare.py | 409 | # Step 3: Restore original ref | COMMENT |
| LOW⚡ | scripts/run_benchmark_compare.py | 413 | # Step 4: Print comparison | COMMENT |
| LOW⚡ | fla/ops/kda/chunk_intra.py | 838 | # Step 1: Run token_parallel first to compute diagonal blocks into Akkd (fp32) | COMMENT |
| LOW⚡ | fla/ops/kda/chunk_intra.py | 839 | # Step 1: compute diagonal blocks into Akk_diag (fp32) | COMMENT |
| LOW | fla/ops/kda/chunk_intra.py | 878 | # Step 2: Fused inter + solve_tril (works for both fixed-len and varlen) | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 120 | # Step 1: compute all 10 lower-triangular [BC, BC] blocks of K @ K^T | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 172 | # Step 2: apply gate and beta scaling | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 215 | # Step 3: forward substitution on diagonal blocks -> (I + A_diag)^{-1} | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 254 | # Step 4: block merge -> full (I + A)^{-1} | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 294 | # Step 5: store full (I + A)^{-1} to output A | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 382 | # Step 1: fused kkt + solve_tril | STRING |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 401 | # Step 1: mathematically equivalent unfused kkt + solve_tril | STRING |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 418 | # Step 2: recompute_w_u | STRING |
| LOW | fla/ops/precond_kda/chunk_intra.py | 547 | # Step 1: Compute diagonal blocks into Akk_diag (fp32) | COMMENT |
| LOW | fla/ops/precond_kda/chunk_intra.py | 585 | # Step 2: Fused inter + solve_tril | COMMENT |
| LOW | fla/ops/precond_kda/chunk_intra.py | 608 | # Step 3: WY representation | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 73 | # Step 1: ATK preconditioning with symmetric fast squash | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 88 | # Step 2: KDA intra-chunk computation (asymmetric) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 105 | # Step 3: Hidden state update (uses kg which is k_precond gated) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 118 | # Step 4: Output computation | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 187 | # Step 1: Recompute WY representation (asymmetric) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 202 | # Step 2: Recompute hidden state (uses kg = gated k_precond) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 232 | # Step 3: dAqk and local dv (matching KDA's chunk_kda_bwd_dAv) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 245 | # Step 4: Backward through hidden state | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 282 | # Step 7: Asymmetric intra backward | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/context_parallel/test_cp_gdp.py | 80 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to all) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdp.py | 114 | # Step 2: Reference Run (recurrent, varlen, no CP) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdp.py | 148 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_gdp.py | 187 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_conv.py | 119 | # Step 1: Prepare Global Data | COMMENT |
| LOW | tests/context_parallel/test_cp_conv.py | 137 | # Step 2: Reference Run | COMMENT |
| LOW | tests/context_parallel/test_cp_conv.py | 161 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_conv.py | 205 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW⚡ | tests/context_parallel/test_cp_token_shift.py | 94 | # Step 1: Prepare Global Data | COMMENT |
| LOW⚡ | tests/context_parallel/test_cp_token_shift.py | 104 | # Step 2: Reference Run | COMMENT |
| LOW⚡ | tests/context_parallel/test_cp_token_shift.py | 114 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_token_shift.py | 141 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_dplr.py | 107 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to all) | COMMENT |
| LOW | tests/context_parallel/test_cp_dplr.py | 147 | # Step 2: Reference Run using dplr_recurrence (ground truth) | COMMENT |
| LOW | tests/context_parallel/test_cp_dplr.py | 224 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_dplr.py | 283 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn2.py | 141 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to other ranks) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn2.py | 214 | # Step 2: Reference Run using naive_recurrent_gdn2 (ground truth) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn2.py | 306 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn2.py | 352 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_kda.py | 141 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to other ranks) | COMMENT |
| LOW | tests/context_parallel/test_cp_kda.py | 210 | # Step 2: Reference Run using naive_recurrent_kda (ground truth) | COMMENT |
| LOW | tests/context_parallel/test_cp_kda.py | 308 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_kda.py | 367 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn.py | 130 | # Step 1: Prepare Global Data (all generated on rank 0, broadcast to all) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn.py | 163 | # Step 2: Reference Run (single GPU, varlen, no CP) | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn.py | 193 | # Step 3: Context Parallel Run | COMMENT |
| LOW | tests/context_parallel/test_cp_gdn.py | 229 | # Step 4: Result Aggregation and Verification | COMMENT |
| LOW | scripts/run_benchmark_compare.py | 371 | # Step 1: HEAD | COMMENT |
| LOW | scripts/run_benchmark_compare.py | 394 | # Step 2: BASE | COMMENT |
| LOW⚡ | scripts/run_benchmark_compare.py | 409 | # Step 3: Restore original ref | COMMENT |
| LOW⚡ | scripts/run_benchmark_compare.py | 413 | # Step 4: Print comparison | COMMENT |
| LOW⚡ | fla/ops/kda/chunk_intra.py | 838 | # Step 1: Run token_parallel first to compute diagonal blocks into Akkd (fp32) | COMMENT |
| LOW⚡ | fla/ops/kda/chunk_intra.py | 839 | # Step 1: compute diagonal blocks into Akk_diag (fp32) | COMMENT |
| LOW | fla/ops/kda/chunk_intra.py | 878 | # Step 2: Fused inter + solve_tril (works for both fixed-len and varlen) | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 120 | # Step 1: compute all 10 lower-triangular [BC, BC] blocks of K @ K^T | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 172 | # Step 2: apply gate and beta scaling | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 215 | # Step 3: forward substitution on diagonal blocks -> (I + A_diag)^{-1} | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 254 | # Step 4: block merge -> full (I + A)^{-1} | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 294 | # Step 5: store full (I + A)^{-1} to output A | COMMENT |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 382 | # Step 1: fused kkt + solve_tril | STRING |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 401 | # Step 1: mathematically equivalent unfused kkt + solve_tril | STRING |
| LOW | fla/ops/gated_delta_rule/chunk_fwd.py | 418 | # Step 2: recompute_w_u | STRING |
| LOW | fla/ops/precond_kda/chunk_intra.py | 547 | # Step 1: Compute diagonal blocks into Akk_diag (fp32) | COMMENT |
| LOW | fla/ops/precond_kda/chunk_intra.py | 585 | # Step 2: Fused inter + solve_tril | COMMENT |
| LOW | fla/ops/precond_kda/chunk_intra.py | 608 | # Step 3: WY representation | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 73 | # Step 1: ATK preconditioning with symmetric fast squash | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 88 | # Step 2: KDA intra-chunk computation (asymmetric) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 105 | # Step 3: Hidden state update (uses kg which is k_precond gated) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 118 | # Step 4: Output computation | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 187 | # Step 1: Recompute WY representation (asymmetric) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 202 | # Step 2: Recompute hidden state (uses kg = gated k_precond) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 232 | # Step 3: dAqk and local dv (matching KDA's chunk_kda_bwd_dAv) | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 245 | # Step 4: Backward through hidden state | COMMENT |
| LOW | fla/ops/precond_kda/chunk.py | 282 | # Step 7: Asymmetric intra backward | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/context_parallel/test_cp_conv.py | 242 | except Exception as e: | CODE |
| LOW | tests/context_parallel/test_cp_token_shift.py | 167 | except Exception as e: | CODE |
| LOW | tests/context_parallel/test_cp_dplr.py | 348 | except Exception as e: | CODE |
| LOW | tests/context_parallel/test_cp_gdn2.py | 416 | except Exception as e: | CODE |
| LOW | tests/context_parallel/test_cp_kda.py | 415 | except Exception as e: | CODE |
| LOW | tests/context_parallel/test_cp_gdn.py | 281 | except Exception as e: | CODE |
| LOW | benchmarks/benchmark_training_throughput.py | 48 | except Exception: | CODE |
| MEDIUM | benchmarks/benchmark_training_throughput.py | 36 | def _git_describe() -> str: | CODE |
| LOW | benchmarks/ops/run.py | 219 | except Exception: | CODE |
| LOW | benchmarks/ops/run.py | 237 | except Exception: | CODE |
| LOW | benchmarks/ops/run.py | 368 | except Exception as e: | CODE |
| LOW | benchmarks/ops/run.py | 383 | except Exception as e: | CODE |
| LOW | benchmarks/ops/run.py | 405 | except Exception as e: | CODE |
| LOW | benchmarks/ops/run.py | 613 | except Exception as e: | CODE |
| LOW | benchmarks/ops/run.py | 656 | except Exception as e: | CODE |
| LOW | benchmarks/ops/run.py | 781 | except Exception as e: | CODE |
| LOW | benchmarks/ops/graph_benchmark.py | 158 | except Exception: | CODE |
| LOW | benchmarks/cp/test_gdn_with_cp.py | 57 | except Exception: | CODE |
| LOW | scripts/utils/autotune_export.py | 62 | except Exception as e: | CODE |
| MEDIUM | scripts/utils/autotune_export.py | 63 | print(f"Error processing {autotune_file}: {e}") | CODE |
| LOW | scripts/utils/autotune_export.py | 107 | except Exception: | CODE |
| LOW | scripts/utils/autotune_export.py | 206 | except Exception: | CODE |
| LOW | scripts/utils/autotune_export.py | 303 | except Exception as e: | CODE |
| MEDIUM | scripts/utils/autotune_export.py | 304 | print(f"Error saving {output_file}: {e}") | CODE |
| LOW | fla/backends.py | 98 | except Exception as error: | CODE |
| LOW⚡ | fla/utils/hardware.py | 34 | except Exception: | CODE |
| LOW⚡ | fla/utils/hardware.py | 40 | except Exception: | CODE |
| LOW | fla/utils/hardware.py | 61 | except Exception: | CODE |
| LOW | fla/utils/hardware.py | 72 | except Exception: | CODE |
| LOW | fla/utils/hardware.py | 156 | except Exception: | CODE |
| LOW | fla/utils/hardware.py | 181 | except Exception: | CODE |
| MEDIUM | fla/utils/hardware.py | 58 | def get_available_device() -> str: | CODE |
| MEDIUM | fla/utils/hardware.py | 150 | def get_all_max_shared_mem(): | CODE |
| LOW | fla/utils/ascend_ub_manager.py | 38 | except Exception: | CODE |
| LOW | fla/utils/ascend_ub_manager.py | 187 | except Exception: | CODE |
| LOW | fla/utils/ascend_ub_manager.py | 231 | except Exception as e: | CODE |
| LOW | …eralized_delta_rule/dplr/backends/tilelang/__init__.py | 211 | except Exception as exc: | CODE |
| LOW | fla/ops/utils/cache.py | 259 | except Exception as e: | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | tests/models/test_cache.py | 421 | attn_state=attn_state, | COMMENT |
| LOW | tests/ops/test_gdn2.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | tests/modules/test_layernorm.py | 321 | assert_close(' dw', ref_dw, tri_dw, 1e-3) | COMMENT |
| LOW | utils/convert_from_llama.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | utils/convert_from_rwkv6.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | .github/workflows/ascend-a2-benchmark-ci.yml | 41 | return; | COMMENT |
| LOW | .github/workflows/oj.yml | 1 | name: oj | COMMENT |
| LOW | fla/layers/gdn2.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/models/mamba/modeling_mamba.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/models/mamba/configuration_mamba.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/models/mamba2/configuration_mamba2.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/models/mamba2/modeling_mamba2.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/models/mamba2/modeling_mamba2.py | 221 | # Reinitialize selected weights subject to the OpenAI GPT-2 Paper Scheme: | COMMENT |
| LOW | fla/ops/titans/naive.py | 81 | # Calculate n_{i,j} | COMMENT |
| LOW | fla/ops/titans/naive.py | 101 | # f[..., t] += (beta[..., t] / beta[..., i]) * m[..., i] | COMMENT |
| LOW | fla/ops/titans/log_impl.py | 41 | # f[..., t] += torch.exp(log_beta[..., t] - log_beta[..., i] + log_m[..., i]) | COMMENT |
| LOW | fla/ops/moba/__init__.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/hgrn/chunk.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/atk/__init__.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/simple_gla/backends/__init__.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/simple_gla/backends/triton_ascend/__init__.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/utils/constant.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/attn/backends/__init__.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/gdn2/chunk_intra_token_parallel.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/gdn2/chunk_intra.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/gdn2/naive.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/ops/gdn2/wy_fast.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/modules/fused_bitlinear.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/modules/norm/layernorm_quant.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/modules/norm/layernorm.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/modules/norm/layernorm_gated.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/modules/norm/triton_ascend/__init__.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| LOW | fla/modules/causal_conv1d/backends/__init__.py | 1 | # Copyright (c) 2023-2026, Songlin Yang, Yu Zhang, Zhiyuan Li | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | scripts/utils/autotune_export.py | 206 | except Exception: | CODE |
| HIGH | fla/utils/ascend_ub_manager.py | 38 | except Exception: | CODE |
| HIGH | fla/utils/ascend_ub_manager.py | 187 | except Exception: | CODE |
| HIGH | fla/utils/ascend_ub_manager.py | 206 | except ValueError: | CODE |
| HIGH | fla/ops/attnres/backends/__init__.py | 19 | except ImportError: | CODE |
| HIGH | fla/modules/causal_conv1d/backends/triton_ascend.py | 36 | except ImportError: | CODE |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | fla/ops/kda/gate.py | 8 | # This file is modified and supported by the Moonshot AI Team | COMMENT |
| MEDIUM⚡ | fla/ops/titans/naive.py | 30 | # Create a matrix filled with 1s for cumulative product | COMMENT |
| MEDIUM⚡ | fla/ops/titans/naive.py | 37 | # Create an expanded version of theta | COMMENT |
| MEDIUM⚡ | fla/ops/titans/naive.py | 39 | # Create a mask to keep only the upper triangular part (excluding the diagonal) | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | evals/ppl.py | 236 | # Print results | COMMENT |
| LOW | tests/conftest.py | 73 | # Check if this frame is from a test file | COMMENT |
| LOW | legacy/training/flame/data.py | 210 | # Check if padding is necessary. | COMMENT |
| LOW | benchmarks/cp/benchmark_kda_cp8_vs_cp2tp.py | 420 | # Print results | COMMENT |
| LOW | fla/utils/env.py | 44 | # Check if triton-windows is installed | COMMENT |
| LOW | fla/models/rwkv7/modeling_rwkv7.py | 340 | # Check if the layer contains the old 'x_x' parameter | COMMENT |
| LOW | fla/ops/utils/cache.py | 80 | # Check if GPU name is overridden via environment variable | COMMENT |
| LOW | fla/ops/utils/cache.py | 82 | # Check if GPU name is overridden via environment variable | COMMENT |
| LOW | fla/ops/utils/cache.py | 106 | # Check if custom config dir is set via environment variable | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | fla/layers/moba.py | 211 | # defines a seamless switch from block-sparse to full attention for the | COMMENT |
| MEDIUM | fla/ops/cp/context.py | 147 | # 2. Operate on CPU to avoid D2H sync and leverage vectorization (int64/long) | COMMENT |
| LOW | fla/modules/feature_map.py | 182 | # currently we simply use linear layers without any non-linear activations | STRING |
| LOW | fla/modules/norm/layernorm.py | 1322 | # us to simply just use from_local | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | tests/ops/test_gdn_kernels.py | 789 | # compensated by the orchestration's reverse cumsum (see chunk_gated_delta_rule_bwd), | COMMENT |
| MEDIUM | fla/ops/gdn2/chunk_fwd.py | 8 | # Forward orchestration for GDN-2 chunkwise training. | COMMENT |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| HIGH | fla/ops/attnres/backends/triton_ascend/fused.py | 125 | The ``BN>1 && BD < D`` D-loop is untested (A2 never reaches it; bwd | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| LOW | fla/utils/decorators.py | 236 | def my_function(do_reduce_labels): | STRING |
| LOW | fla/utils/decorators.py | 246 | def my_function(max_size): | STRING |
| Severity | File | Line | Snippet | Context |
|---|---|---|---|---|
| MEDIUM | fla/ops/abc/naive.py | 82 | This is just for demonstration purposes, with no numerical stabilities guaranteed. | STRING |