xax 0.1.12__py3-none-any.whl → 0.1.14__py3-none-any.whl

xax/__init__.py

@@ -12,7 +12,7 @@ and running the update script:
     python -m scripts.update_api --inplace
 """
 
-__version__ = "0.1.12"
+__version__ = "0.1.14"
 
 # This list shouldn't be modified by hand; instead, run the update script.
 __all__ = [
@@ -75,7 +75,9 @@ __all__ = [
     "Task",
     "collate",
     "collate_non_null",
+    "breakpoint_if_nan",
     "get_named_leaves",
+    "log_if_nan",
     "BaseFileDownloader",
     "ContextTimer",
     "CumulativeTimer",
@@ -234,7 +236,9 @@ NAME_MAP: dict[str, str] = {
     "Task": "task.task",
     "collate": "utils.data.collate",
     "collate_non_null": "utils.data.collate",
+    "breakpoint_if_nan": "utils.debugging",
     "get_named_leaves": "utils.debugging",
+    "log_if_nan": "utils.debugging",
     "BaseFileDownloader": "utils.experiments",
     "ContextTimer": "utils.experiments",
     "CumulativeTimer": "utils.experiments",
@@ -386,7 +390,7 @@ if IMPORT_ALL or TYPE_CHECKING:
     from xax.task.script import Script, ScriptConfig
     from xax.task.task import Config, Task
     from xax.utils.data.collate import CollateMode, collate, collate_non_null
-    from xax.utils.debugging import get_named_leaves
+    from xax.utils.debugging import breakpoint_if_nan, get_named_leaves, log_if_nan
     from xax.utils.experiments import (
         BaseFileDownloader,
         ContextTimer,

xax/task/base.py

@@ -82,9 +82,6 @@ class BaseTask(Generic[Config]):
     def on_after_checkpoint_save(self, ckpt_path: Path, state: State) -> State:
         return state
 
-    def on_before_checkpoint_load(self, ckpt_path: Path) -> None:
-        pass
-
     @functools.cached_property
     def task_class_name(self) -> str:
         return self.__class__.__name__

xax/task/loggers/stdout.py

@@ -14,7 +14,7 @@ from xax.utils.text import Color, colored, format_timedelta
 
 def format_number(value: int | float, precision: int) -> str:
     if isinstance(value, int):
-        return str(value)
+        return f"{value:,}"  # Add commas to the number
     return f"{value:.{precision}g}"
 
 
@@ -80,11 +80,10 @@ class StdoutLogger(LoggerImpl):
         self.write_fp.write("\033[2J\033[H")
 
     def write_state_window(self, line: LogLine) -> None:
-        elapsed_time = format_timedelta(datetime.timedelta(seconds=line.state.elapsed_time_s), short=True)
-        state_info = {
-            "Steps": f"{line.state.num_steps}",
-            "Samples": f"{line.state.num_samples}",
-            "Elapsed Time": f"{elapsed_time}",
+        state_info: dict[str, str] = {
+            "Steps": format_number(line.state.num_steps, 0),
+            "Samples": format_number(line.state.num_samples, 0),
+            "Elapsed Time": format_timedelta(datetime.timedelta(seconds=line.state.elapsed_time_s), short=True),
         }
 
         colored_prefix = colored("Phase: ", "grey", bold=True)

xax/task/mixins/checkpointing.py

@@ -98,19 +98,39 @@ class CheckpointingMixin(ArtifactsMixin[Config], Generic[Config]):
     ) -> tuple[PyTree, State, DictConfig]: ...
 
     @overload
-    def load_checkpoint(self, path: Path, part: Literal["model"]) -> PyTree: ...
+    def load_checkpoint(
+        self,
+        path: Path,
+        part: Literal["model"],
+    ) -> PyTree: ...
 
     @overload
-    def load_checkpoint(self, path: Path, part: Literal["opt"]) -> optax.GradientTransformation: ...
+    def load_checkpoint(
+        self,
+        path: Path,
+        part: Literal["opt"],
+    ) -> optax.GradientTransformation: ...
 
     @overload
-    def load_checkpoint(self, path: Path, part: Literal["opt_state"]) -> optax.OptState: ...
+    def load_checkpoint(
+        self,
+        path: Path,
+        part: Literal["opt_state"],
+    ) -> optax.OptState: ...
 
     @overload
-    def load_checkpoint(self, path: Path, part: Literal["state"]) -> State: ...
+    def load_checkpoint(
+        self,
+        path: Path,
+        part: Literal["state"],
+    ) -> State: ...
 
     @overload
-    def load_checkpoint(self, path: Path, part: Literal["config"]) -> DictConfig: ...
+    def load_checkpoint(
+        self,
+        path: Path,
+        part: Literal["config"],
+    ) -> DictConfig: ...
 
     def load_checkpoint(
         self,
@@ -125,9 +145,6 @@ class CheckpointingMixin(ArtifactsMixin[Config], Generic[Config]):
         | State
         | DictConfig
     ):
-        # Calls the base callback.
-        self.on_before_checkpoint_load(path)
-
         with tarfile.open(path, "r:gz") as tar:
 
             def get_model() -> PyTree:

xax/task/mixins/compile.py

@@ -32,6 +32,10 @@ def get_cache_dir() -> str | None:
 @dataclass
 class CompileOptions:
     # JAX compilation options
+    debug_nans: bool = field(
+        value=False,
+        help="If True, breaks on NaNs",
+    )
     disable_jit: bool = field(
         value=False,
         help="If True, disables JIT compilation",
@@ -89,6 +93,10 @@ class CompileMixin(BaseTask[Config], Generic[Config]):
         cc = self.config.compile
 
         # Set basic compilation flags
+        if cc.debug_nans:
+            logger.info("Enabling NaNs debugging")
+            jax.config.update("jax_debug_nans", True)
+
         if cc.disable_jit:
             logger.info("Disabling JIT compilation")
             jax.config.update("jax_disable_jit", True)

xax/utils/debugging.py

@@ -4,6 +4,8 @@ from collections import deque
 from collections.abc import Iterable, Mapping
 from typing import Any, Callable, Deque
 
+import jax
+import jax.numpy as jnp
 from jaxtyping import Array
 
 
@@ -47,3 +49,11 @@ def get_named_leaves(
                     q.append((depth + 1, gname, cnode))
 
     return ret
+
+
+def breakpoint_if_nan(x: Array) -> None:
+    jax.lax.cond(jnp.any(jnp.isnan(x)), lambda: jax.debug.breakpoint(), lambda: None)
+
+
+def log_if_nan(x: Array, loc: str) -> None:
+    jax.lax.cond(jnp.any(jnp.isnan(x)), lambda: jax.debug.print("=== NaNs: {loc} ===", loc=loc), lambda: None)

{xax-0.1.12.dist-info → xax-0.1.14.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xax
-Version: 0.1.12
+Version: 0.1.14
 Summary: A library for fast Jax experimentation
 Home-page: https://github.com/kscalelabs/xax
 Author: Benjamin Bolte

{xax-0.1.12.dist-info → xax-0.1.14.dist-info}/RECORD

@@ -1,4 +1,4 @@
-xax/__init__.py,sha256=7vdTYO7jAJdDxKZURlFxc3Y5kr5mVQcTQjeh_sYjD6I,13834
+xax/__init__.py,sha256=D7czvfKKQJlemPuatMPVYbAO4ST3U272QRIyTOru7JI,13989
 xax/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 xax/requirements-dev.txt,sha256=qkscNkFzWd1S5fump-AKH53rR65v2x5FmboFdy_kKvs,128
 xax/requirements.txt,sha256=9LAEZ5c5gqRSARRVA6xJsVTa4MebPZuC4yOkkwkZJFw,297
@@ -16,7 +16,7 @@ xax/nn/norm.py,sha256=WgZ3QCrUnf-YecwhEtVPcr99fKK3ECl_UeiAs2uv7oo,564
 xax/nn/parallel.py,sha256=fnTiT7MsG7eQrJvqwjIz2Ifo3P27TuxIJzmpGYSa_dQ,4608
 xax/nn/ssm.py,sha256=8dLAcQ1hBaMT-kkHvwGu_ecxJeTY32WeMYmd4T4KtxA,10745
 xax/task/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-xax/task/base.py,sha256=E4l1yCrAkM2TVTbVYrmk6BoVHMkbD4IYsTT921XOyi0,7760
+xax/task/base.py,sha256=DqgGIlo5kEWpYix3DdPCEkCgVLUOocjyFr8okaSUq-k,7680
 xax/task/logger.py,sha256=1SZjVC6UCtZUoMPcpp3ckotL324QDeYDvHVhf5MHVqg,36271
 xax/task/script.py,sha256=bMMIJoUtpSBvPp6-7bejTrajTXvSg0794sYLKdPIToE,972
 xax/task/task.py,sha256=UHMpnv__gqMcfbC_L-Hhk-DCnUYlFVsgbNf-v8o8B7U,1424
@@ -28,12 +28,12 @@ xax/task/loggers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,
 xax/task/loggers/callback.py,sha256=lyuZX6Bir7xJM07ifdQIl1jlclgkiS82UO9V4y7wgPs,1582
 xax/task/loggers/json.py,sha256=yXHb1bmfsEnk4p0F1Up1ertWYdcPAFZm25NT8wE3Jb8,4045
 xax/task/loggers/state.py,sha256=6bG-NRsSUzAukYiglCT0oDj8zRMpffH4e1TKWGw1x4k,959
-xax/task/loggers/stdout.py,sha256=bR0k-PfmFgLfPxLPb4hZw_8G_msA32UeHfAAu11nEYs,6757
+xax/task/loggers/stdout.py,sha256=BBXqr95gNt5KuCN8XyKnTJF8JdwkR4JgLKrkvcaTBVM,6788
 xax/task/loggers/tensorboard.py,sha256=kI8LvBuBBhPgkP8TeaTQb9SQ0FqaIodwQh2SuWDCnIA,7706
 xax/task/mixins/__init__.py,sha256=D3oU31rB9FeOr9MPLleLt5JFbftUr4sBTwgnwQdc2qA,809
 xax/task/mixins/artifacts.py,sha256=2ezmZGzPGe3nhsd9KRkeHWWXdbT9m7drzimIfw6v1XY,2892
-xax/task/mixins/checkpointing.py,sha256=a6tVyISsDIz68rrhb1rAh3rjQlqkDVJCmSBmETQrnRM,8480
-xax/task/mixins/compile.py,sha256=8jEdlGs-a14N_CwZA3Rxe461MT83dyIDr3Z56VkjviQ,3693
+xax/task/mixins/checkpointing.py,sha256=nRddgtasagf0oTZE9LE5IN5JY7jy4BD_M0rlqYp4sCM,8554
+xax/task/mixins/compile.py,sha256=PG5aF3W9v_xGiImHgUJ7gmwuQQoSQWufdpl2N_mlLX0,3922
 xax/task/mixins/cpu_stats.py,sha256=C_t71UTrv4LwQzhO5iubsfomj4jYa9bzpE4zBcHdoHM,9211
 xax/task/mixins/data_loader.py,sha256=WjMWk9uACfBMMClLMcLPkE0WNIvlCZnmqyyqLqJpjX0,6545
 xax/task/mixins/gpu_stats.py,sha256=IGPBro9xzSivwD43zM18lWcuei7IhA8LilxSPHqNl4I,8747
@@ -43,7 +43,7 @@ xax/task/mixins/runnable.py,sha256=IYIsLd2k09g-_y6o44EhJqT7E6BpsyEMmsyLSuzqjtc,1
 xax/task/mixins/step_wrapper.py,sha256=-Yu5Nft2CRw1JvZt6J_94SM1vqX8fk08IDK95Pmd2ew,1648
 xax/task/mixins/train.py,sha256=aIebtOIvERYofSyqzNGBpNYlNrXweqFUqM9dHiTx3Dc,26253
 xax/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-xax/utils/debugging.py,sha256=9WlCrEqbq-SVXPEM4rhsLYERH97XNX7XSYLSI3sgKGk,1619
+xax/utils/debugging.py,sha256=OtUdu-3tQsQtik0Q9UM-SNV46IbPjwrAfZcywzoB5d4,1940
 xax/utils/experiments.py,sha256=5CUja1H_cx4dnVqTGQekOpIhqISwHtAgLxZ34GV7cwM,29229
 xax/utils/jax.py,sha256=tC0NNelbrSTzwNGluiwLGKtoHhVpgdzrv-xherB3VtY,4752
 xax/utils/jaxpr.py,sha256=S80nyEkv188RInzq3kCAdkQCU-bf6s0oPTrCE_LjkRs,2298
@@ -58,8 +58,8 @@ xax/utils/data/collate.py,sha256=Rd9vMomr_S_zCa_Hi4dO-8ntzAfVwndIUtuXFA3iNcc,706
 xax/utils/types/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 xax/utils/types/frozen_dict.py,sha256=ZCMGfSfr2_b2qZbq9ywPD0zej5tpVSId2JftXpwfB5k,4686
 xax/utils/types/hashable_array.py,sha256=l5iIcFmkYzfGeaZmcSoeFkthFASqM8xJYK3AXhZQYwc,992
-xax-0.1.12.dist-info/licenses/LICENSE,sha256=HCN2bImAzUOXldAZZI7JZ9PYq6OwMlDAP_PpX1HnuN0,1071
-xax-0.1.12.dist-info/METADATA,sha256=hLRAX5__7QjBgjzhxbRftGvEsNrt8IAdgd22dMtHu_Y,1878
-xax-0.1.12.dist-info/WHEEL,sha256=CmyFI0kx5cdEMTLiONQRbGQwjIoR1aIYB7eCAQ4KPJ0,91
-xax-0.1.12.dist-info/top_level.txt,sha256=g4Au_r2XhvZ-lTybviH-Fh9g0zF4DAYHYxPue1-xbs8,4
-xax-0.1.12.dist-info/RECORD,,
+xax-0.1.14.dist-info/licenses/LICENSE,sha256=HCN2bImAzUOXldAZZI7JZ9PYq6OwMlDAP_PpX1HnuN0,1071
+xax-0.1.14.dist-info/METADATA,sha256=WbKtAXJUYKHvBrOJPEm_eXF9O9ekc0WdPmsQQCSGG5Q,1878
+xax-0.1.14.dist-info/WHEEL,sha256=CmyFI0kx5cdEMTLiONQRbGQwjIoR1aIYB7eCAQ4KPJ0,91
+xax-0.1.14.dist-info/top_level.txt,sha256=g4Au_r2XhvZ-lTybviH-Fh9g0zF4DAYHYxPue1-xbs8,4
+xax-0.1.14.dist-info/RECORD,,