nshtrainer 0.8.4__py3-none-any.whl → 0.8.6__py3-none-any.whl

nshtrainer/model/base.py

@@ -15,7 +15,7 @@ from lightning.fabric.utilities.types import _MAP_LOCATION_TYPE, _PATH
 from lightning.pytorch import LightningDataModule, LightningModule, Trainer
 from lightning.pytorch.callbacks import Callback
 from lightning.pytorch.utilities.types import STEP_OUTPUT
-from typing_extensions import Self, TypeVar, deprecated, override
+from typing_extensions import Self, TypeVar, override
 
 from .config import (
     BaseConfig,
@@ -39,12 +39,6 @@ THparams = TypeVar("THparams", bound=BaseConfig, infer_variance=True)
 
 
 class Base(DebugModuleMixin, Generic[THparams]):
-    @deprecated("Use `ll.nn.MLP` instead.")
-    def mlp(self, *args, **kwargs):
-        from ..nn.mlp import MLP
-
-        return MLP(*args, **kwargs)
-
     @torch.jit.unused
     @property
     def config(self) -> THparams:

nshtrainer/trainer/signal_connector.py

@@ -150,7 +150,7 @@ class _SignalConnector(_LightningSignalConnector):
         cmd = ["scontrol", "requeue", job_id]
 
         # requeue job
-        log.info(f"requeing job {job_id}...")
+        log.info(f"Requeuing job {job_id}...")
         try:
             result = subprocess.call(cmd)
         except FileNotFoundError:
@@ -202,18 +202,46 @@ class _SignalConnector(_LightningSignalConnector):
             exe = str((Path(bin_dir) / exe).resolve().absolute())
 
         log.info(f"Using LSF requeue executable: {exe}")
-        cmd = [exe, job_id]
-
-        # Requeue job
-        log.info(f"Requeuing job {job_id}...")
-        try:
-            result = subprocess.call(cmd)
-        except FileNotFoundError:
-            # Retry with shell context if subprocess call fails
-            result = subprocess.call(" ".join(cmd), shell=True)
 
-        # Print result text
-        if result == 0:
-            log.info(f"Requeued LSF job: {job_id}")
+        # If NSHRUNNER_LSF_EXIT_SCRIPT_DIR exists, we should emit a bash script in that directory
+        # rather than calling the requeue command directly. This is because the requeue command
+        # is only available outside of the `jsrun` context, and the exit script is called within
+        # the `jsrun` context.
+        if not (exit_script_dir := os.getenv("NSHRUNNER_LSF_EXIT_SCRIPT_DIR")):
+            cmd = [exe, job_id]
+
+            # Requeue job
+            log.info(f"Requeuing job {job_id}...")
+            try:
+                result = subprocess.call(cmd)
+            except FileNotFoundError:
+                # Retry with shell context if subprocess call fails
+                result = subprocess.call(" ".join(cmd), shell=True)
+
+            # Print result text
+            if result == 0:
+                log.info(f"Requeued LSF job: {job_id}")
+            else:
+                log.warning(
+                    f"Requeuing LSF job {job_id} failed with error code {result}"
+                )
         else:
-            log.warning(f"Requeuing LSF job {job_id} failed with error code {result}")
+            log.critical(
+                "Environment variable NSHRUNNER_LSF_EXIT_SCRIPT_DIR found.\n"
+                "Writing requeue script to exit script directory."
+            )
+            exit_script_dir = Path(exit_script_dir)
+            assert (
+                exit_script_dir.is_dir()
+            ), f"Exit script directory {exit_script_dir} does not exist"
+
+            exit_script_path = exit_script_dir / f"requeue_{job_id}.sh"
+            log.info(f"Writing requeue script to {exit_script_path}")
+
+            with exit_script_path.open("w") as f:
+                f.write(f"#!/bin/bash\n{exe} {job_id}\n")
+
+            # Make the script executable
+            os.chmod(exit_script_path, 0o755)
+
+            log.info(f"Requeue script written to {exit_script_path}")

{nshtrainer-0.8.4.dist-info → nshtrainer-0.8.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: nshtrainer
-Version: 0.8.4
+Version: 0.8.6
 Summary: 
 Author: Nima Shoghi
 Author-email: nimashoghi@gmail.com

{nshtrainer-0.8.4.dist-info → nshtrainer-0.8.6.dist-info}/RECORD

@@ -46,7 +46,7 @@ nshtrainer/lr_scheduler/_base.py,sha256=7xOIuxQ86YHbFWG5a3gX46emQj1WN_LaY4-i0Q1T
 nshtrainer/lr_scheduler/linear_warmup_cosine.py,sha256=mn6cyizyI_stkXtg6zxIEGF9btIxMRWigUHUTlUYCSw,5221
 nshtrainer/lr_scheduler/reduce_lr_on_plateau.py,sha256=h76oTHYpMxauV_l6lviya5DW-WKArwxxf7ZQizhmbCw,2782
 nshtrainer/model/__init__.py,sha256=y32Hla-5whpzLL2BtCJpBakSp8o-1nQbpO0j_-xq_Po,1864
-nshtrainer/model/base.py,sha256=k2Qk_rH4FsBOib2PISXnG97SBlgeD1w_jw02NE4zEQI,21460
+nshtrainer/model/base.py,sha256=YtqnjiMf0cLVjFEQuOLm5WwCkVnZftiHlIdCrxdax3s,21297
 nshtrainer/model/config.py,sha256=6lATW6-Z1SIDgQ1IWrGBVQKTr8DhL5b_rFbJHQz0d5o,66796
 nshtrainer/model/modules/callback.py,sha256=JF59U9-CjJsAIspEhTJbVaGN0wGctZG7UquE3IS7R8A,6408
 nshtrainer/model/modules/debug.py,sha256=DTVty8cKnzj1GCULRyGx_sWTTsq9NLi30dzqjRTnuCU,1127
@@ -65,13 +65,13 @@ nshtrainer/runner.py,sha256=7EumpnBkdNWjSNT9Gm-pkxAJ3W6-iMC-yae-WNeZcLw,3771
 nshtrainer/scripts/check_env.py,sha256=IMl6dSqsLYppI0XuCsVq8lK4bYqXwY9KHJkzsShz4Kg,806
 nshtrainer/scripts/find_packages.py,sha256=FbdlfmAefttFSMfaT0A46a-oHLP_ioaQKihwBfBeWeA,1467
 nshtrainer/trainer/__init__.py,sha256=P2rmr8oBVTHk-HJHYPcUwWqDEArMbPR4_rPpATbWK3E,40
-nshtrainer/trainer/signal_connector.py,sha256=QAoPM_C5JJOVQebcrJOimUUD3GHyoeZUqCEAvzZlT4U,8710
+nshtrainer/trainer/signal_connector.py,sha256=WlB7yB7d6F-dKyJEcdsxxkpYq4lWVSl2OoiZKlFAuaw,10001
 nshtrainer/trainer/trainer.py,sha256=eYEYfY9v70MuorHcSf8nqM7f2CkmUHhpPcjCk4FJD7k,14034
 nshtrainer/util/environment.py,sha256=_SEtiQ_s5bL5pllUlf96AOUv15kNvCPvocVC13S7mIk,4166
 nshtrainer/util/seed.py,sha256=HEXgVs-wldByahOysKwq7506OHxdYTEgmP-tDQVAEkQ,287
 nshtrainer/util/slurm.py,sha256=rofIU26z3SdL79SF45tNez6juou1cyDLz07oXEZb9Hg,1566
 nshtrainer/util/typed.py,sha256=NGuDkDzFlc1fAoaXjOFZVbmj0mRFjsQi1E_hPa7Bn5U,128
 nshtrainer/util/typing_utils.py,sha256=8ptjSSLZxlmy4FY6lzzkoGoF5fGNClo8-B_c0XHQaNU,385
-nshtrainer-0.8.4.dist-info/METADATA,sha256=PS0Z7icSp22Ok5Bs6SnOc0PEBqg42kjXTnpW5jqQaxw,647
-nshtrainer-0.8.4.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-nshtrainer-0.8.4.dist-info/RECORD,,
+nshtrainer-0.8.6.dist-info/METADATA,sha256=5BOWb0DXYCmqYGIzVTApTBj9wOoo9V8so3yfRuYzE6g,647
+nshtrainer-0.8.6.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+nshtrainer-0.8.6.dist-info/RECORD,,