| Files | Last commit | Last update |
|---|---|---|
style: clear platform static-scan findings No behavior change: verified per file with before/after differential runs (HEAD vs working tree) and the repo's own tools (pylint 3.3.7, lizard 1.24.0). - trainer/config/resolver.py: split the container/class branches out of coerce_value (CCN 19 -> 11); declare __all__ before imports. - trainer/config/manager.py: merge the override helpers back into _replace_target_path/_replace_path and inline the single-use temporaries (locals 17 -> 14/12), keep the shared _suggestion helper; declare __all__ before imports; read the callable through the public attribute. - trainer/config/target.py: expose the wrapped callable as a public read-only property so callers stop touching the protected _target_. - trainer/config/parallelism.py: drop the unused build_options imports. - trainer/config/__init__.py, models/_transformers/config_resolver.py, components/losses/masked_ce.py: declare __all__ before imports. - auto_parallel/config_adapter/_config_loader.py: restore the second blank line before a top-level definition. - auto_parallel/config_adapter/_strategy_output.py: use dict.get() for the nested accelerator lookup instead of double subscripting. - trainer/runtime/metrics.py: drop the stale FIXME line before the docstring. | 1 month ago | |
refactor!: reorganize auto_models into top-level models/components/distributed/trainer/data Squashed 58 commits of the AutoModels directory restructure (M0-M5, S4a-S8d, top-level split): - characterization tests first (M0): UT collection baseline, public API contract snapshots, build-order golden cases, shared fixtures, training YAML snapshots - distributed split (S4a-S4g): sharding_config/planner/applier/injection split into distributed/{plan,recipe_spec,apply,compile, activation_checkpoint} + private _builder/; CP/EP into context_parallel/ and expert_parallel/; fsdp2 into fsdp_adapter + source_shard - models adapter layer (M1-M5): adapter_spec/registry skeleton, Qwen3-MoE replacements and CP/EP contracts migrated into models/qwen3_moe/adapter, single recipe converged onto models/qwen3_moe/recipes/train.yaml, legacy fusion files deleted; per-family sharding_rules declared via ModelAdapterSpec with directory-convention auto-discovery registration - build pipeline (S5a-S5g): checkpoint finalization into checkpoint_loader, plan DTO boundary normalized, god module split into model_builder, registry split into config_resolver + models/registry - data stack (S6a-S6f): indexed/text/online/batching/parallel/vlm/tools migrated to top-level hyper_parallel/data - trainer (S7a-S7e): config tree/callbacks/runtime/base+text+vlm migrated to top-level hyper_parallel/trainer, utils split eliminated - closeout (S8a-S8d): legacy core/shard API, legacy hyper_parallel/models and components compatibility package deleted; import boundary gates - top-level split: auto_models/ dissolved into models/ (adapter layer + api/build_options/replacement/_transformers, external interface), components/ (modules/functional/quantization/losses/optim/checkpoint) and distributed/; references rewritten across py/yaml/sh and setup.py; boundary gates updated (distributed -> models limited to the build_options leaf DTO plus the planner's function-level registry lookup) Verified: full UT 4562 passed / 39 failed (identical to the pre-refactor environment-failure baseline), characterization suites green. | 2 months ago | |
style: clear platform static-scan findings No behavior change: verified per file with before/after differential runs (HEAD vs working tree) and the repo's own tools (pylint 3.3.7, lizard 1.24.0). - trainer/config/resolver.py: split the container/class branches out of coerce_value (CCN 19 -> 11); declare __all__ before imports. - trainer/config/manager.py: merge the override helpers back into _replace_target_path/_replace_path and inline the single-use temporaries (locals 17 -> 14/12), keep the shared _suggestion helper; declare __all__ before imports; read the callable through the public attribute. - trainer/config/target.py: expose the wrapped callable as a public read-only property so callers stop touching the protected _target_. - trainer/config/parallelism.py: drop the unused build_options imports. - trainer/config/__init__.py, models/_transformers/config_resolver.py, components/losses/masked_ce.py: declare __all__ before imports. - auto_parallel/config_adapter/_config_loader.py: restore the second blank line before a top-level definition. - auto_parallel/config_adapter/_strategy_output.py: use dict.get() for the nested accelerator lookup instead of double subscripting. - trainer/runtime/metrics.py: drop the stale FIXME line before the docstring. | 1 month ago | |
clean code | 1 month ago | |
style: clear platform static-scan findings No behavior change: verified per file with before/after differential runs (HEAD vs working tree) and the repo's own tools (pylint 3.3.7, lizard 1.24.0). - trainer/config/resolver.py: split the container/class branches out of coerce_value (CCN 19 -> 11); declare __all__ before imports. - trainer/config/manager.py: merge the override helpers back into _replace_target_path/_replace_path and inline the single-use temporaries (locals 17 -> 14/12), keep the shared _suggestion helper; declare __all__ before imports; read the callable through the public attribute. - trainer/config/target.py: expose the wrapped callable as a public read-only property so callers stop touching the protected _target_. - trainer/config/parallelism.py: drop the unused build_options imports. - trainer/config/__init__.py, models/_transformers/config_resolver.py, components/losses/masked_ce.py: declare __all__ before imports. - auto_parallel/config_adapter/_config_loader.py: restore the second blank line before a top-level definition. - auto_parallel/config_adapter/_strategy_output.py: use dict.get() for the nested accelerator lookup instead of double subscripting. - trainer/runtime/metrics.py: drop the stale FIXME line before the docstring. | 1 month ago | |
style: clear platform static-scan findings No behavior change: verified per file with before/after differential runs (HEAD vs working tree) and the repo's own tools (pylint 3.3.7, lizard 1.24.0). - trainer/config/resolver.py: split the container/class branches out of coerce_value (CCN 19 -> 11); declare __all__ before imports. - trainer/config/manager.py: merge the override helpers back into _replace_target_path/_replace_path and inline the single-use temporaries (locals 17 -> 14/12), keep the shared _suggestion helper; declare __all__ before imports; read the callable through the public attribute. - trainer/config/target.py: expose the wrapped callable as a public read-only property so callers stop touching the protected _target_. - trainer/config/parallelism.py: drop the unused build_options imports. - trainer/config/__init__.py, models/_transformers/config_resolver.py, components/losses/masked_ce.py: declare __all__ before imports. - auto_parallel/config_adapter/_config_loader.py: restore the second blank line before a top-level definition. - auto_parallel/config_adapter/_strategy_output.py: use dict.get() for the nested accelerator lookup instead of double subscripting. - trainer/runtime/metrics.py: drop the stale FIXME line before the docstring. | 1 month ago | |
style: clear platform static-scan findings No behavior change: verified per file with before/after differential runs (HEAD vs working tree) and the repo's own tools (pylint 3.3.7, lizard 1.24.0). - trainer/config/resolver.py: split the container/class branches out of coerce_value (CCN 19 -> 11); declare __all__ before imports. - trainer/config/manager.py: merge the override helpers back into _replace_target_path/_replace_path and inline the single-use temporaries (locals 17 -> 14/12), keep the shared _suggestion helper; declare __all__ before imports; read the callable through the public attribute. - trainer/config/target.py: expose the wrapped callable as a public read-only property so callers stop touching the protected _target_. - trainer/config/parallelism.py: drop the unused build_options imports. - trainer/config/__init__.py, models/_transformers/config_resolver.py, components/losses/masked_ce.py: declare __all__ before imports. - auto_parallel/config_adapter/_config_loader.py: restore the second blank line before a top-level definition. - auto_parallel/config_adapter/_strategy_output.py: use dict.get() for the nested accelerator lookup instead of double subscripting. - trainer/runtime/metrics.py: drop the stale FIXME line before the docstring. | 1 month ago | |
refactor!: reorganize auto_models into top-level models/components/distributed/trainer/data Squashed 58 commits of the AutoModels directory restructure (M0-M5, S4a-S8d, top-level split): - characterization tests first (M0): UT collection baseline, public API contract snapshots, build-order golden cases, shared fixtures, training YAML snapshots - distributed split (S4a-S4g): sharding_config/planner/applier/injection split into distributed/{plan,recipe_spec,apply,compile, activation_checkpoint} + private _builder/; CP/EP into context_parallel/ and expert_parallel/; fsdp2 into fsdp_adapter + source_shard - models adapter layer (M1-M5): adapter_spec/registry skeleton, Qwen3-MoE replacements and CP/EP contracts migrated into models/qwen3_moe/adapter, single recipe converged onto models/qwen3_moe/recipes/train.yaml, legacy fusion files deleted; per-family sharding_rules declared via ModelAdapterSpec with directory-convention auto-discovery registration - build pipeline (S5a-S5g): checkpoint finalization into checkpoint_loader, plan DTO boundary normalized, god module split into model_builder, registry split into config_resolver + models/registry - data stack (S6a-S6f): indexed/text/online/batching/parallel/vlm/tools migrated to top-level hyper_parallel/data - trainer (S7a-S7e): config tree/callbacks/runtime/base+text+vlm migrated to top-level hyper_parallel/trainer, utils split eliminated - closeout (S8a-S8d): legacy core/shard API, legacy hyper_parallel/models and components compatibility package deleted; import boundary gates - top-level split: auto_models/ dissolved into models/ (adapter layer + api/build_options/replacement/_transformers, external interface), components/ (modules/functional/quantization/losses/optim/checkpoint) and distributed/; references rewritten across py/yaml/sh and setup.py; boundary gates updated (distributed -> models limited to the build_options leaf DTO plus the planner's function-level registry lookup) Verified: full UT 4562 passed / 39 failed (identical to the pre-refactor environment-failure baseline), characterization suites green. | 2 months ago | |
refactor!: reorganize auto_models into top-level models/components/distributed/trainer/data Squashed 58 commits of the AutoModels directory restructure (M0-M5, S4a-S8d, top-level split): - characterization tests first (M0): UT collection baseline, public API contract snapshots, build-order golden cases, shared fixtures, training YAML snapshots - distributed split (S4a-S4g): sharding_config/planner/applier/injection split into distributed/{plan,recipe_spec,apply,compile, activation_checkpoint} + private _builder/; CP/EP into context_parallel/ and expert_parallel/; fsdp2 into fsdp_adapter + source_shard - models adapter layer (M1-M5): adapter_spec/registry skeleton, Qwen3-MoE replacements and CP/EP contracts migrated into models/qwen3_moe/adapter, single recipe converged onto models/qwen3_moe/recipes/train.yaml, legacy fusion files deleted; per-family sharding_rules declared via ModelAdapterSpec with directory-convention auto-discovery registration - build pipeline (S5a-S5g): checkpoint finalization into checkpoint_loader, plan DTO boundary normalized, god module split into model_builder, registry split into config_resolver + models/registry - data stack (S6a-S6f): indexed/text/online/batching/parallel/vlm/tools migrated to top-level hyper_parallel/data - trainer (S7a-S7e): config tree/callbacks/runtime/base+text+vlm migrated to top-level hyper_parallel/trainer, utils split eliminated - closeout (S8a-S8d): legacy core/shard API, legacy hyper_parallel/models and components compatibility package deleted; import boundary gates - top-level split: auto_models/ dissolved into models/ (adapter layer + api/build_options/replacement/_transformers, external interface), components/ (modules/functional/quantization/losses/optim/checkpoint) and distributed/; references rewritten across py/yaml/sh and setup.py; boundary gates updated (distributed -> models limited to the build_options leaf DTO plus the planner's function-level registry lookup) Verified: full UT 4562 passed / 39 failed (identical to the pre-refactor environment-failure baseline), characterization suites green. | 2 months ago |
| Files | Last commit | Last update |
|---|---|---|
| 1 month ago | ||
| 2 months ago | ||
| 1 month ago | ||
| 1 month ago | ||
| 1 month ago | ||
| 1 month ago | ||
| 1 month ago | ||
| 2 months ago | ||
| 2 months ago |