aquin.compute.simulate_inputs Namespace Reference

Functions
Path	_expand (str path)
dict[str, Any]	_normalize_row (dict[str, Any] row)
list[dict]	load_dataset_file (str path)
Any|None	_parse_structured_file (Path p)
None	_collect_hints (Any obj, dict[str, Any] found, int depth=0)
dict[str, Any]	_try_training_hints (Path path)
dict	prepare_simulation_args (dict args)

Variables
dict	_HINT_KEYS
dict	_KEY_ALIASES
tuple	_INSTR_FIELDS = ("instruction", "prompt", "input", "query", "question", "text")
tuple	_RESP_FIELDS = ("response", "output", "answer", "completion", "target", "label")

Function Documentation

_collect_hints()

None _collect_hints ( Any obj, dict[str, Any] found, int depth = 0 )

protected

Definition at line 133 of file simulate_inputs.py.

References _collect_hints().

Referenced by _collect_hints(), and _try_training_hints().

_expand()

Path _expand ( str path )

protected

Definition at line 42 of file simulate_inputs.py.

Referenced by load_dataset_file(), and prepare_simulation_args().

_normalize_row()

dict[str, Any] _normalize_row ( dict[str, Any] row )

protected

Definition at line 46 of file simulate_inputs.py.

Referenced by load_dataset_file().

_parse_structured_file()

Any | None _parse_structured_file ( Path p )

protected

Definition at line 114 of file simulate_inputs.py.

Referenced by _try_training_hints().

_try_training_hints()

dict[str, Any] _try_training_hints ( Path path )

protected

Best-effort scrape of rank/lr/etc. from the user's existing config :  never required.

Definition at line 148 of file simulate_inputs.py.

References _collect_hints(), and _parse_structured_file().

Referenced by prepare_simulation_args().

load_dataset_file()

list[dict] load_dataset_file

(

str

path

)

Load rows from the user's dataset file (.json, .jsonl, .csv).

Definition at line 61 of file simulate_inputs.py.

References _expand(), and _normalize_row().

Referenced by prepare_simulation_args().

prepare_simulation_args()

dict prepare_simulation_args

(

dict

args

)

Resolve --dataset and --algo paths from the user's existing pipeline.
No Aquin-specific config format :  paths only. Optional CLI flags override hints.

Definition at line 158 of file simulate_inputs.py.

References _expand(), _try_training_hints(), and load_dataset_file().

Variable Documentation

_HINT_KEYS

dict aquin.compute.simulate_inputs._HINT_KEYS

protected

Initial value:

=  {
    "rank", "alpha", "lr", "learning_rate", "epochs", "num_train_epochs",
    "dropout", "target_modules", "targetModules", "warmup_steps", "warmupSteps",
    "grad_clip", "gradClip", "weight_decay", "weightDecay", "grad_accum_steps",
    "gradAccumSteps", "optimizer", "scheduler", "max_seq_len", "maxSeqLen",
    "use_qlora", "useQlora", "use_rlhf", "rlhf_beta",
}

Definition at line 19 of file simulate_inputs.py.

_INSTR_FIELDS

tuple aquin.compute.simulate_inputs._INSTR_FIELDS = ("instruction", "prompt", "input", "query", "question", "text")

protected

Definition at line 38 of file simulate_inputs.py.

_KEY_ALIASES

dict aquin.compute.simulate_inputs._KEY_ALIASES

protected

Initial value:

=  {
    "learning_rate": "lr",
    "num_train_epochs": "epochs",
    "target_modules": "targetModules",
    "warmup_steps": "warmupSteps",
    "grad_clip": "gradClip",
    "weight_decay": "weightDecay",
    "grad_accum_steps": "gradAccumSteps",
    "max_seq_len": "maxSeqLen",
    "use_qlora": "useQlora",
}

Definition at line 26 of file simulate_inputs.py.

_RESP_FIELDS

tuple aquin.compute.simulate_inputs._RESP_FIELDS = ("response", "output", "answer", "completion", "target", "label")

protected

Definition at line 39 of file simulate_inputs.py.