[BUG] Silent crash in safetensors call when compiling shards

[r0mar0ma]

OS

Windows

GPU Library

CUDA 12.x

Python version

3.10

Pytorch version

torch-2.7.0+cu128

Model

No response

Describe the bug

Crashing silently in processing source safetensors on CPU when compiling shards.
safetensors-0.5.3

Reproduction steps

Run conversion with -cf

Expected behavior

Should compile shards and complete conversion successfully

Logs

No response

Additional context

At least on my config moving safe_open to GPU (RTX4090) is resolving a problem.
Here is just a draft example with first GPU (device = 0):

compile.py

- with safe_open(file, framework = "pt") as f:
-     tensor = f.get_tensor(lkey)

+ with safe_open(file, framework = "pt", device = 0) as f:
+     tensor = f.get_tensor(lkey).to("cpu")

Acknowledgements

• I have looked for similar issues before submitting this one.
• I understand that the developers have lives and my issue will be answered when possible.
• I understand the developers of this program are human, and I will ask my questions politely.