huggingface/safetensors

Issues

• Documentation about multipart safetensors

  #532 opened 2 months ago by attashe
  0

• Can't save checkpoint to file descriptor.

  #531 opened 2 months ago by GLivshits
  0

• let `safetensors.load` support load weight from bytearray

  #529 opened 2 months ago by JeryZeng
  0

• RuntimeError: unable to mmap 23802932552 bytes from file <./flux1-dev.safetensors>: Cannot allocate memory (12)

  #528 opened 2 months ago by sakura-nyaa
  0

• [Question] Comparison with the zarr format?

  #527 opened 2 months ago by julioasotodv
  0

• Retrieving saved metadata

  #521 opened 3 months ago by parlance-zz
  3

• Wrong device when using device="cpu" with torch.device

  #503 opened 4 months ago by chrisdesa
  3

• add remapping to safetensors.torch.load_model

  #524 opened 3 months ago by f-fuchs
  0

• SafetensorError does not appear in `__init__.pyi`

  #523 opened 3 months ago by alanhdu
  0

• Some layers are not loaded(‘encoder.embed_token.weight’ and 'shared.weight') when loading converted safetensors into transfomrers.MT5EncoderModel.

  #520 opened 3 months ago by WongGawa
  0

• converting weight of more than 50G‘s .bin to .safetensors failed

  #519 opened 3 months ago by WongGawa
  0

• Problem serializing quantized weights

  #515 opened 4 months ago by sayakpaul
  3

• inconsistent behavior of slicing in certain model file

  #495 opened 4 months ago by youkaichao
  4

• misleading error message in `safe_open`

  #485 opened 4 months ago by Clement-Lelievre
  4

• pytorch: safetensors library hardcodes using CUDA if only device index is provided

  #499 opened 4 months ago by dvrogozh
  3

• SafetensorError is not pickleable, potentially causing hangs in multi-process environments

  #492 opened 4 months ago by Wehrdo
  2

• Failed to load `.safetensors` as state dict with error from `torch.frombuffer` in `safetensors.torch.load`

  #442 opened 4 months ago by RunDevelopment
  10

• How to save model checkpoint from a distributed training from multiple nodes?

  #490 opened 4 months ago by Emerald01
  2

• Allow serialization of TensorInfo, and other safetensors classes

  #463 opened 8 months ago by LeoDog896
  7

• Compilation fail due to test error

  #491 opened 4 months ago by BananPrzydawka
  3

• Load incorrect element values in bfloat16 tensor on big-endian

  #448 opened 4 months ago by kiszk
  3

• Unable to save llama3 since complex64 is not supported

  #489 opened 4 months ago by npuichigo
  1

• Register a MIME type for the `safetensors` format.

  #486 opened 4 months ago by judahrand
  1

• README is inaccurate

  #487 opened 4 months ago by judahrand
  3

• Poor performance for reading Numpy

  #460 opened 4 months ago by ZHUI
  11

• get_tensor returns an empty tensor on LoRa safetensors adapter saved by SFTTrainer

  #498 opened 4 months ago by tsvisab
  1

• get_slice has the same speed as get_tensor

  #496 opened 4 months ago by youkaichao
  1

• KeyError: torch.complex64 when attempting to save PyTorch model

  #450 opened 6 months ago by NielsRogge
  4

• Support multi-device loading for load_file() with pt framework

  #477 opened 5 months ago by mobicham
  2

• Error when using map_location="cuda:0" with multiple GPUs

  #481 opened 5 months ago by AAnirudh07
  1

• safetensors.tensorflow.save modifies its input argument

  #480 opened 5 months ago by allaisandrea
  1

• Can't seem to skip parameter initialization while using the `safetensors.torch.load_model` API!

  #478 opened 5 months ago by goelayu
  1

• Official C++ API lib/API

  #475 opened 6 months ago by dharmeshkakadia
  1

• Inaccurate performance benchmarks

  #476 opened 6 months ago by goelayu
  1

• ModuleNotFoundError: No module named 'torch._higher_order_ops'

  #482 opened 6 months ago by guotong1988
  1

• Request for writable temporary directory handling within Python tests

  #471 opened 6 months ago by salayatana66
  1

• error saving AI model weights with non-contiguous tensors

  #469 opened 6 months ago by not-lain
  1

• Incorrect serialization given only metadata

  #466 opened 6 months ago by tommyip
  1

• Typing error using with safe_open

  #459 opened 7 months ago by chanind
  1

• Safetensors uses excessive RAM when saving files

  #458 opened 7 months ago by sheepymeh
  1

• How to fully load checkpointed weights in memory?

  #474 opened 7 months ago by goelayu
  1

• Extend data type support for MLX's metadata

  #457 opened 7 months ago by mzbac
  1

• benchmark results discussion for safetensor and Pytorch native

  #455 opened 7 months ago by Jeffwan
  1

• Efficient key-wise streaming

  #443 opened 7 months ago by ljleb
  4

• Saving individual tensors to file

  #447 opened 8 months ago by mar-muel
  1

• Add support for saving to `_ZipWriteFile`

  #454 opened 8 months ago by stancld
  0

• Loading a model with safetensors results in OOM error whereas torch.load() does not.

  #437 opened 8 months ago by maxdebayser
  1

• Ruby bindings

  #445 opened 8 months ago by ankane
  0

• Slice loading errors with integer indices

  #439 opened 9 months ago by hct-synth
  3

• load_state_dict doesnt support torch._subclasses.fake_tensor.FakeTensorMode

  #441 opened 9 months ago by thiagocrepaldi
  1