File size: 27,077 Bytes
1bfc3f0 ce32ca3 1bfc3f0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 | """Action-space-bound ActionCodec2Sparse frontend for independent action chunks."""
from __future__ import annotations
import json
import sys
import warnings
from collections.abc import Mapping
from pathlib import Path
from typing import Any, TextIO
import numpy as np
import torch
from transformers.feature_extraction_utils import BatchFeature
from ...backends.actioncodec2 import ActionCodec2Config, PresetConfig
from ...backends.actioncodec2.defaults import builtin_action_spaces
from ...backends.actioncodec2.hf import (
GENERATED_CARD_MARKER,
PROCESSOR_CONFIG_NAME,
resolve_pretrained_directory,
)
from ...backends.actioncodec2.preprocessing import PreprocessingSpec
from ...backends.actioncodec2.release import MANIFEST_NAME
from ...exporting import save_runtime
from ...backends.actioncodec2.semantic import resampled_horizon
from ...backends.actioncodec2.sparse import (
ActionCodec2SparseProcessor,
ActionCodec2SparseRouter,
SparseLayout,
SparseProfileTrainer,
SparseVocabulary,
sparse_profile_config,
)
from ...backends.actioncodec2.sparse.routing import SPARSE_ROUTER_CONFIG_NAME, save_router_metadata
from ...backends.actioncodec2.training import build_primitives
from ..conversion import (
_numpy,
_numpy_token_rows,
restore_source_layout,
)
from ..base_action_tokenizer import BaseActionTokenizer
from ..layout import ActionSpaceMixin
from ..training import TrainingMixin
from .loading import describe_release, load_release_router
ALLOW_PATTERNS = [
SPARSE_ROUTER_CONFIG_NAME,
PROCESSOR_CONFIG_NAME,
"config.json",
MANIFEST_NAME,
"profiles/**",
"joint/**",
"eef/**",
"fit_report.json",
"README.md",
]
_ASSETS = Path(__file__).absolute().parents[2] / "backends" / "actioncodec2" / "assets"
def _profile_config(raw) -> ActionCodec2Config:
if isinstance(raw, ActionCodec2Config):
return raw
if isinstance(raw, Mapping):
return ActionCodec2Config.from_dict(raw)
return ActionCodec2Config.from_yaml(raw)
def build_fresh_router(
profile_configs=None, primitive_order: int = 1, vocab_size: int = 4096
) -> ActionCodec2SparseRouter:
"""Unfitted joint/EEF profiles; defaults derive from the packaged profiles of that order."""
if type(primitive_order) is not int or primitive_order not in (1, 2):
raise ValueError("primitive_order must be 1 or 2")
if type(vocab_size) is not int or vocab_size < 1:
raise ValueError("vocab_size must be a positive integer")
if profile_configs is not None and vocab_size != 4096:
raise ValueError("set vocabulary budgets in profile_configs when providing custom profiles")
configs = dict(profile_configs or {})
if configs.keys() - {"joint", "eef"}:
raise ValueError("profile_configs keys must be 'joint' or 'eef'")
suffix = "" if primitive_order == 1 else "_second_order"
processors, budgets = {}, {}
for name in ("joint", "eef"):
if name in configs:
config = _profile_config(configs[name])
else:
default = ActionCodec2Config.from_yaml(_ASSETS / "profiles" / f"{name}{suffix}.yaml")
config = sparse_profile_config(default, vocab_size=vocab_size)
if config.train is None or config.train.sparse is None:
raise ValueError(f"{name} profile config requires train.sparse")
primitives = build_primitives(config)
layout = SparseLayout.from_config(config, primitives)
vocabulary = SparseVocabulary(layout, (), (), config.set_bpe.lookahead)
processors[name] = ActionCodec2SparseProcessor(config, primitives, vocabulary)
budgets[name] = config.train.bpe.vocab_size
return ActionCodec2SparseRouter.from_processors(
processors, token_budgets=budgets, presets=list(builtin_action_spaces().values())
)
class ActionCodec2Sparse(ActionSpaceMixin, TrainingMixin, BaseActionTokenizer):
"""ActionCodec2 physics with a zero-free token layer, bound to one action space."""
trainer_class = SparseProfileTrainer
model_type = "actioncodec2_sparse"
encode_kwargs = frozenset({"fps", "current_state", "previous_first_order"})
def __init__(
self,
backend: ActionCodec2SparseRouter | None = None,
*,
action_space: str,
primitive_order: int = 1,
vocab_size: int = 4096,
profile_configs: Mapping[str, object] | None = None,
) -> None:
fresh = backend is None
if fresh:
backend = build_fresh_router(profile_configs, primitive_order, vocab_size)
elif profile_configs is not None or primitive_order != 1 or vocab_size != 4096:
raise ValueError("profile construction options cannot accompany an existing backend")
if not isinstance(backend, ActionCodec2SparseRouter):
raise TypeError(
f"ActionCodec2Sparse requires ActionCodec2SparseRouter, got {type(backend).__name__}"
)
preset = backend.preset(action_space)
if preset.target_hz != 15.0:
raise ValueError("ActionCodec2Sparse artifacts require target_hz=15")
self.action_tokenizer = backend
self._action_space = preset
self._action_vocab_size = int(backend.vocab_size)
self._new_action_tokens = backend.new_action_tokens
self.action_dim = preset.input_dim
self.requires_previous_first_order = bool(backend.second_order_boundaries(preset))
self.max_action_token_len = None
self.fixed_action_token_len = None
self._warned_default_fps = False
super().__init__(append_bar_sentinels=False)
self._fitted_profiles = frozenset() if fresh else frozenset(backend.profiles)
self.training_report = None
self.preprocessing: dict[str, PreprocessingSpec] = {}
self.release_metadata: dict[str, Any] | None = None
def __repr__(self) -> str:
return (
f"ActionCodec2Sparse(action_space={self.action_space.name!r}, "
f"D={self.action_dim}, vocab={self.vocab_size})"
)
def _describe_action_spaces(self, action_dim: int | None = None) -> str:
fitted = ", ".join(sorted(self.fitted_profiles)) or "none; call codec.fit(episodes)"
layouts = self.action_tokenizer.describe_action_spaces(
selected=self.action_space.name, action_dim=action_dim
)
return f"Fitted profiles: {fitted}\n{layouts}"
@staticmethod
def builtin_action_spaces():
"""Return the immutable named layout definitions shipped in the package."""
return builtin_action_spaces()
def print_action_spaces(self, *, action_dim: int | None = None, file: TextIO | None = None):
"""Print all layouts or candidates matching ``action_dim``."""
print(self._describe_action_spaces(action_dim), file=sys.stdout if file is None else file)
def for_action_space(self, name: str) -> ActionCodec2Sparse:
"""Share current weights with a new binding."""
codec = type(self)(self.action_tokenizer, action_space=name)
codec._fitted_profiles = self.fitted_profiles
codec.training_report = self.training_report
codec.preprocessing = dict(self.preprocessing)
codec.release_metadata = self.release_metadata
return codec
@property
def fitted_profiles(self) -> frozenset[str]:
"""Names of profiles fitted on data; persisted by save_pretrained()."""
return self._fitted_profiles
@property
def is_fitted(self) -> bool:
"""Whether every profile used by the selected action space is fitted."""
return all(c.space in self.fitted_profiles for c in self.action_space.components)
def _require_fitted(self) -> None:
missing = {c.space for c in self.action_space.components} - self.fitted_profiles
if missing:
raise RuntimeError(
f"Unfitted profiles {sorted(missing)}; call codec.fit(episodes) first"
)
def _source_fps(self, fps, preset=None) -> float:
preset = self.action_space if preset is None else preset
if fps is None:
if not self._warned_default_fps:
warnings.warn(
f"fps was omitted; assuming {preset.source_hz:g} Hz from action_space "
f"{preset.name!r}. Pass the actual fps explicitly to avoid changing action timing.",
UserWarning,
stacklevel=3,
)
self._warned_default_fps = True
return float(preset.source_hz)
return float(fps)
def codec_horizon(self, horizon: int, *, fps: float | None = None) -> int:
"""Codec frames (15 Hz) of a ``horizon``-frame source chunk at ``fps``."""
if isinstance(horizon, bool) or not isinstance(horizon, int) or horizon < 1:
raise ValueError("horizon must be a positive integer")
return resampled_horizon(
horizon, source_hz=self._source_fps(fps), target_hz=self.action_space.target_hz
)
def max_tokens_for_horizon(self, horizon: int, *, fps: float | None = None) -> int:
"""Upper bound on the tokens of one encoded ``horizon``-frame source chunk."""
frames = self.codec_horizon(horizon, fps=fps)
components = self.action_space.components
width = sum(len(c.active_dims) + 1 for c in components)
return frames * width + len(components) - 1
def grammar(self, horizon: int, *, fps: float | None = None):
"""Constrained-generation grammar for one ``horizon``-frame source chunk."""
self._require_fitted()
return self.action_tokenizer.grammar(
preset=self.action_space.name, horizon=self.codec_horizon(horizon, fps=fps)
)
def logits_processor(
self, horizon, *, fps, prompt_length, eos_token_id, token_offset=0, pad_token_id=None
):
"""Create a Transformers LogitsProcessor for one ``horizon``-frame chunk."""
from .generation import ActionCodec2SparseLogitsProcessor
return ActionCodec2SparseLogitsProcessor(
self,
horizon,
fps=fps,
prompt_length=prompt_length,
eos_token_id=eos_token_id,
token_offset=token_offset,
pad_token_id=pad_token_id,
)
def _check_state(self, current_state, action: str):
name = self.action_space.name
if self.action_space.input == "absolute" and current_state is None:
raise ValueError(
f"action space {name!r} is absolute; {action} requires current_state with shape [B, D]"
)
if self.action_space.input == "delta" and current_state is not None:
raise ValueError(f"action space {name!r} is delta; omit current_state")
return _numpy(current_state)
def encode(
self, action, *, fps=None, current_state=None, previous_first_order=None, encode_kwargs=None
) -> list[list[int]]:
"""Encode ``(B,T,D)`` or ``(T,D)`` source chunks into ragged token rows."""
kwargs = self._check_encode_kwargs(encode_kwargs)
direct = {
"fps": fps,
"current_state": current_state,
"previous_first_order": previous_first_order,
}
duplicates = sorted(k for k, v in direct.items() if v is not None and k in kwargs)
if duplicates:
raise ValueError(f"encode arguments provided twice: {duplicates}")
kwargs.update({k: v for k, v in direct.items() if v is not None})
values = torch.as_tensor(action)
if values.ndim == 2:
values = values.unsqueeze(0)
return self._encode_action_indices(values, encode_kwargs=kwargs)
def __call__(
self,
action,
*,
fps=None,
current_state=None,
previous_first_order=None,
encode_kwargs=None,
return_tensors=None,
) -> BatchFeature:
rows = self.encode(
action,
fps=fps,
current_state=current_state,
previous_first_order=previous_first_order,
encode_kwargs=encode_kwargs,
)
return BatchFeature({"input_ids": rows}, tensor_type=return_tensors)
def _encode_action_indices(self, action, encode_kwargs=None):
self._require_fitted()
kwargs = dict(encode_kwargs or {})
current_state = self._check_state(kwargs.get("current_state"), "encode")
values = _numpy(action)
if values.ndim != 3 or values.shape[2] != self.action_dim:
raise ValueError(f"actions must have shape [B, T_source, {self.action_dim}]")
return self.action_tokenizer.encode_batch(
values,
preset=self.action_space.name,
fps=self._source_fps(kwargs.get("fps")),
current_state=current_state,
previous_first_order=self._numpy_boundary(kwargs.get("previous_first_order")),
)
def decode(
self,
action_token_ids,
*,
horizon: int,
fps=None,
current_state=None,
previous_first_order=None,
) -> torch.Tensor:
"""Decode rows of one ``horizon``-frame source chunk to a CPU ``(B,T,D)`` tensor."""
return self.decode_token_ids_to_actions_batch(
action_token_ids,
time_horizon=horizon,
decode_kwargs={
"fps": fps,
"current_state": current_state,
"previous_first_order": previous_first_order,
},
)
def decode_token_ids_to_actions_batch(
self, action_token_ids, *, time_horizon=None, action_dim=None, decode_kwargs=None
) -> torch.Tensor:
"""Decode integer rows (ragged or ``(B,L)``) of one source ``time_horizon``."""
rows = [
self._checked_decode_row(row).tolist() for row in _numpy_token_rows(action_token_ids)
]
actions, _ = self._decode_action_indices(
rows, time_horizon=time_horizon, action_dim=action_dim, decode_kwargs=decode_kwargs
)
return torch.from_numpy(actions)
def _decode_action_indices(
self, action_indices, *, time_horizon=None, action_dim=None, decode_kwargs=None
):
self._require_fitted()
if time_horizon is None:
raise ValueError("ActionCodec2Sparse decode requires the chunk horizon")
if action_dim is not None and int(action_dim) != self.action_dim:
raise ValueError(
f"action_dim={action_dim} disagrees with action space D={self.action_dim}"
)
kwargs = dict(decode_kwargs or {})
unknown = sorted(set(kwargs) - self.encode_kwargs)
if unknown:
raise ValueError(f"ActionCodec2Sparse does not consume decode_kwargs {unknown}")
fps = self._source_fps(kwargs.get("fps"))
decoded = self.action_tokenizer.decode_batch(
action_indices,
preset=self.action_space.name,
horizon=self.codec_horizon(time_horizon, fps=fps),
fps=fps,
current_state=self._check_state(kwargs.get("current_state"), "decode"),
previous_first_order=self._numpy_boundary(kwargs.get("previous_first_order")),
)
actions = restore_source_layout(self.action_space, decoded)
return actions, [set() for _ in range(len(actions))]
def batch_decode(
self,
token_rows,
*,
horizon: int,
fps=None,
current_state=None,
previous_first_order=None,
errors="raise",
):
"""Decode independent rows of one horizon, isolating invalid rows."""
if errors not in ("raise", "return"):
raise ValueError("errors must be 'raise' or 'return'")
rows = list(token_rows)
current = _numpy(current_state)
if current is not None and np.shape(current) != (len(rows), self.action_dim):
raise ValueError(f"current_state must have shape {(len(rows), self.action_dim)}")
previous = self._numpy_boundary(previous_first_order)
if previous is not None and any(len(v) != len(rows) for v in previous.values()):
raise ValueError(
"previous_first_order arrays must have batch dimension equal to token rows"
)
result = []
for index, row in enumerate(rows):
try:
state = None if current is None else current[index : index + 1]
boundary = (
None
if previous is None
else {k: v[index : index + 1] for k, v in previous.items()}
)
result.append(
self.decode(
[row],
horizon=horizon,
fps=fps,
current_state=state,
previous_first_order=boundary,
)
)
except ValueError as error:
error = ValueError(f"token row {index}: {error}")
if errors == "raise":
raise error from None
result.append(error)
return result
def decode_prefix(
self,
action_token_ids,
*,
horizon: int,
steps: int,
fps=None,
current_state=None,
previous_first_order=None,
):
"""Decode the first ``steps`` 15 Hz frames of streaming prefixes to ``(B,steps,D)``."""
self._require_fitted()
fps = self._source_fps(fps)
if fps != self.action_space.target_hz:
raise ValueError("prefix decode requires fps equal to the 15 Hz codec clock")
frames = self.codec_horizon(horizon, fps=fps)
if isinstance(steps, bool) or not isinstance(steps, int) or not 1 <= steps <= frames:
raise ValueError("steps must be an integer within the chunk horizon")
rows = [self._checked_decode_row(row).tolist() for row in action_token_ids]
grammar = self.action_tokenizer.grammar(preset=self.action_space.name, horizon=frames)
if any(grammar.covered_steps(row) < steps for row in rows):
raise ValueError(f"action prefix does not fix the first {steps} frames")
decoded = self.action_tokenizer.decode_batch(
rows,
preset=self.action_space.name,
horizon=frames,
fps=fps,
current_state=self._check_state(_numpy(current_state), "decode"),
previous_first_order=self._numpy_boundary(previous_first_order),
)
return torch.from_numpy(restore_source_layout(self.action_space, decoded)[:, :steps])
def final_first_order(
self,
action_token_ids,
*,
horizon: int,
previous_first_order,
fps=None,
executed_steps: int | None = None,
) -> dict[str, np.ndarray]:
"""Second-order component boundaries ``{component: (B,D_active)}`` for the next chunk."""
self._require_fitted()
return self.action_tokenizer.final_first_order(
[self._checked_decode_row(row).tolist() for row in action_token_ids],
preset=self.action_space.name,
horizon=self.codec_horizon(horizon, fps=fps),
previous_first_order=self._numpy_boundary(previous_first_order),
executed_steps=executed_steps,
)
@classmethod
def describe_pretrained(
cls,
pretrained_model_name_or_path,
*,
action_dim=None,
gripper_mode="2bin",
primitive_order: int = 1,
**kwargs,
) -> str:
"""Report a release's or saved artifact's layouts from manifests alone."""
directory = resolve_pretrained_directory(
pretrained_model_name_or_path, allow_patterns=ALLOW_PATTERNS, **kwargs
)
if (directory / MANIFEST_NAME).is_file():
return describe_release(
directory,
action_dim=action_dim,
gripper_mode=gripper_mode,
primitive_order=primitive_order,
)
specs, presets = ActionCodec2SparseRouter.read_manifest(directory)
metadata = _saved_metadata(directory)
fitted = ", ".join(metadata.get("fitted_profiles", [s.name for s in specs]))
from ...backends.actioncodec2.presets import format_action_spaces
layouts = format_action_spaces(
{p.name: p for p in presets},
selected=metadata.get("action_space"),
action_dim=action_dim,
codec_name="ActionCodec2Sparse",
chunk_horizon_decode=True,
)
return f"Fitted profiles: {fitted}\n{layouts}"
@classmethod
def from_pretrained(
cls,
pretrained_model_name_or_path,
*,
action_space=None,
gripper_mode: str | None = None,
primitive_order: int | None = None,
verify_gates: bool = False,
**kwargs,
) -> ActionCodec2Sparse:
"""Load a release package or saved frontend and bind an action space."""
if cls is ActionCodec2Sparse and ("variant" in kwargs or (Path(pretrained_model_name_or_path) / "variants.json").is_file()
or (Path(pretrained_model_name_or_path) / "artifact_manifest.json").is_file()):
from ...catalog import ActionCodec2 as Catalog
return Catalog.from_pretrained(pretrained_model_name_or_path, action_space=action_space, **kwargs)
directory = resolve_pretrained_directory(
pretrained_model_name_or_path, allow_patterns=ALLOW_PATTERNS, **kwargs
)
metadata = _saved_metadata(directory)
action_space = metadata.get("action_space") if action_space is None else action_space
if action_space is None:
raise ValueError(
"action_space is required; inspect layouts with "
"print(ActionCodec2Sparse.describe_pretrained(path))"
)
if (directory / MANIFEST_NAME).is_file():
router, preprocessing, release = load_release_router(
directory,
gripper_mode=gripper_mode or "2bin",
primitive_order=1 if primitive_order is None else primitive_order,
verify_gates=verify_gates,
)
codec = cls(router, action_space=_register(router, action_space))
codec.preprocessing, codec.release_metadata = preprocessing, release
return codec
if gripper_mode is not None or primitive_order is not None or verify_gates:
raise ValueError(
"gripper_mode, primitive_order and verify_gates select release variants only"
)
router = ActionCodec2SparseRouter.from_pretrained(directory)
codec = cls(router, action_space=_register(router, action_space))
fitted = metadata.get("fitted_profiles", list(router.profiles))
if not isinstance(fitted, list) or set(fitted) - router.profiles.keys():
raise ValueError("invalid fitted_profiles in processor_config.json")
codec._fitted_profiles = frozenset(fitted)
codec.preprocessing = {
name: PreprocessingSpec.from_dict(spec)
for name, spec in metadata.get("preprocessing", {}).items()
}
codec.release_metadata = metadata.get("release")
report = directory / "fit_report.json"
if report.is_file():
codec.training_report = json.loads(report.read_text(encoding="utf-8"))
return codec
def to_dict(self) -> dict[str, object]:
"""Lightweight discovery metadata saved as ``processor_config.json``."""
payload = {
"action_space": self.action_space.name,
"fitted_profiles": sorted(self.fitted_profiles),
}
if self.preprocessing:
payload["preprocessing"] = {k: v.to_dict() for k, v in self.preprocessing.items()}
return payload
def save_pretrained(
self, save_directory: str | Path, *, include_training_report=False
) -> list[str]:
"""Save inference assets; include the local fit report only when explicitly requested."""
directory = Path(save_directory)
files = self.action_tokenizer.save_pretrained(directory)
files.extend(map(str, save_router_metadata(directory, **self.to_dict())))
files.append(str(_save_model_card(directory, self)))
report = directory / "fit_report.json"
if include_training_report and self.training_report is not None:
report.write_text(json.dumps(self.training_report, indent=2) + "\n", encoding="utf-8")
files.append(str(report))
elif report.exists():
report.unlink()
files.extend(save_runtime(directory))
return files
def eval(self) -> ActionCodec2Sparse:
return self
def _save_model_card(directory: Path, codec: ActionCodec2Sparse) -> Path:
path = directory / "README.md"
if path.exists() and GENERATED_CARD_MARKER not in path.read_text(encoding="utf-8"):
return path
context = 'context = {"fps": 15.0}\n'
if codec.action_space.input == "absolute":
context += 'context["current_state"] = current_state # observed (B, D) source state\n'
if codec.requires_previous_first_order:
context += 'context["previous_first_order"] = previous_first_order # canonical 15 Hz increments by component\n'
path.write_text(
"---\nlibrary_name: actioncodec2\nlicense: apache-2.0\ntags:\n- robotics\n- action-tokenizer\n---\n\n"
f"{GENERATED_CARD_MARKER}\n\n# ActionCodec2Sparse\n\n"
"Install actioncodec2 >= 0.1.0 before loading this artifact.\n\n"
"```python\nfrom actioncodec2 import ActionCodec2Sparse\n\n"
"codec = ActionCodec2Sparse.from_pretrained('./my_codec')\n"
+ context
+ "rows = codec.encode(actions, **context)\n"
"restored = codec.decode(rows, horizon=actions.shape[-2], **context)\n```\n\n"
"Supply actions and any required boundary observations in the declared layout.\n"
"Rows are independent chunks; decode and grammar take the chunk horizon.\n\n"
"## Layouts\n\n```text\n" + str(codec) + "\n```\n",
encoding="utf-8",
)
return path
def _saved_metadata(directory: Path) -> dict:
path = directory / PROCESSOR_CONFIG_NAME
return json.loads(path.read_text(encoding="utf-8")) if path.is_file() else {}
def _register(router: ActionCodec2SparseRouter, action_space) -> str:
if isinstance(action_space, (PresetConfig, Mapping, Path)):
return router.register_action_space(action_space).name
if isinstance(action_space, str):
if Path(action_space).is_file():
return router.register_action_space(Path(action_space)).name
return action_space
raise TypeError("action_space must be a registered name, PresetConfig, mapping, or YAML path")
|