nshtrainer 0.18.1__py3-none-any.whl → 0.18.2__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
nshtrainer/_hf_hub.py CHANGED
@@ -1,3 +1,4 @@
1
+ import io
1
2
  import logging
2
3
  import os
3
4
  from pathlib import Path
@@ -299,7 +300,22 @@ def _save_checkpoint_files(
299
300
  # Resolve the repository name
300
301
  repo_name = _repo_name(api, root_config)
301
302
 
303
+ # Let's read all the files to memory right now,
304
+ # in case they get used/removed by other processes.
305
+ # Read all the files to memory
306
+ file_contents: list[bytes | None] = []
302
307
  for p in paths:
308
+ try:
309
+ with open(p, "rb") as f:
310
+ file_contents.append(f.read())
311
+ except IOError as e:
312
+ log.warning(f"Failed to read checkpoint file {p}: {str(e)}")
313
+ file_contents.append(None)
314
+
315
+ for p, contents in zip(paths, file_contents):
316
+ if contents is None:
317
+ continue
318
+
303
319
  try:
304
320
  relative_path = p.relative_to(checkpoint_dir)
305
321
  except ValueError:
@@ -314,7 +330,7 @@ def _save_checkpoint_files(
314
330
  # Upload the checkpoint file to the repository
315
331
  try:
316
332
  api.upload_file(
317
- path_or_fileobj=str(p.resolve().absolute()),
333
+ path_or_fileobj=io.BytesIO(contents),
318
334
  path_in_repo=str(path_in_repo),
319
335
  repo_id=repo_name,
320
336
  repo_type="model",
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.1
2
2
  Name: nshtrainer
3
- Version: 0.18.1
3
+ Version: 0.18.2
4
4
  Summary:
5
5
  Author: Nima Shoghi
6
6
  Author-email: nimashoghi@gmail.com
@@ -3,7 +3,7 @@ nshtrainer/_checkpoint/loader.py,sha256=myFObRsPdb8jBncMK73vjr5FDJIfKhF86Ec_kSjX
3
3
  nshtrainer/_checkpoint/metadata.py,sha256=p5e7dhVPpOGrXeuesq_7Y_RHi5lguzDAR_UXtMJXzWU,5175
4
4
  nshtrainer/_checkpoint/saver.py,sha256=DkbCH0YeOJ71m32vAARiQdGBf0hvwwdoAV8LOFGy-0Y,1428
5
5
  nshtrainer/_experimental/__init__.py,sha256=pEXPyI184UuDHvfh4p9Kg9nQZQZI41e4_HvNd4BK-yg,81
6
- nshtrainer/_hf_hub.py,sha256=b1Na0-SyOM5xlJCH8cqjk0ggEVCPMI_z770c32JIQRY,10701
6
+ nshtrainer/_hf_hub.py,sha256=Py9_8ADvMCFPaJzeE7bxm8Mgs3mEMkyWJ4pDEccTGt8,11230
7
7
  nshtrainer/callbacks/__init__.py,sha256=4qocBDzQbLLhhbIEfvbA3SQB_Dy9ZJH7keMwPay-ZS8,2359
8
8
  nshtrainer/callbacks/_throughput_monitor_callback.py,sha256=aJo_11rc4lo0IYOd-kHmPDtzdC4ctgXyRudkRJqH4m4,23184
9
9
  nshtrainer/callbacks/actsave.py,sha256=qbnaKts4_dvjPeAaPtv7Ds12_vEWzaHUfg_--49NB9I,4041
@@ -85,6 +85,6 @@ nshtrainer/util/seed.py,sha256=Or2wMPsnQxfnZ2xfBiyMcHFIUt3tGTNeMMyOEanCkqs,280
85
85
  nshtrainer/util/slurm.py,sha256=rofIU26z3SdL79SF45tNez6juou1cyDLz07oXEZb9Hg,1566
86
86
  nshtrainer/util/typed.py,sha256=NGuDkDzFlc1fAoaXjOFZVbmj0mRFjsQi1E_hPa7Bn5U,128
87
87
  nshtrainer/util/typing_utils.py,sha256=8ptjSSLZxlmy4FY6lzzkoGoF5fGNClo8-B_c0XHQaNU,385
88
- nshtrainer-0.18.1.dist-info/METADATA,sha256=kQBOxMbb8io1JEfguuBHEnRkRL2zWXxAYKuhLYrPMnY,935
89
- nshtrainer-0.18.1.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
90
- nshtrainer-0.18.1.dist-info/RECORD,,
88
+ nshtrainer-0.18.2.dist-info/METADATA,sha256=vev96DaxCnqJOAvvGrGOJ37OpWNFLrCdtGPN-kpnvO4,935
89
+ nshtrainer-0.18.2.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
90
+ nshtrainer-0.18.2.dist-info/RECORD,,