[BUG] ImportError when installing on an EC2 instance

[melopeo]

Bug report checklist

• I provided code that demonstrates a minimal reproducible example.
• I confirmed bug exists on the latest mainline of Chronos via source install.

Describe the bug
After installing chronos in editable version, I get the following torch related issue:

import chronos
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "XXXX/python3.9/site-packages/chronos/__init__.py", line 4, in <module>
    from .base import BaseChronosPipeline, ForecastType
  File "XXXX/python3.9/site-packages/chronos/base.py", line 12, in <module>
    import torch
  File "XXXX/python3.9/site-packages/torch/__init__.py", line 367, in <module>
    from torch._C import *  # noqa: F403
ImportError: XXXX/python3.9/site-packages/torch/lib/../../nvidia/cusparse/lib/libcusparse.so.12: undefined symbol: __nvJitLinkComplete_12_4, version libnvJitLink.so.12

One quick way to overcome this is the following

pip uninstall torch
pip install torch --index-url https://download.pytorch.org/whl/cu121

This is potentially related to pytorch/pytorch#134929

Expected behavior
No error shown.

To reproduce
pip install -e .

Environment description
Operating system: Ubuntu 22.04
Python version: 3.9
CUDA version: Build cuda_12.1.r12.1/compiler.32688072_0
PyTorch version: 2.5.1
HuggingFace transformers version: 4.47.1
HuggingFace accelerate version: 0.34.2

[lostella]

The issue appears to have been solved in the latest torch 2.6, #277 enabled that and chronos-forecasting release v1.5.0 includes that. Closing 😎