bump docs to PyTorch 2.5.1 and corresponding nvfuser (#1377)

README.md

@@ -81,11 +81,11 @@ pip install git+https://github.com/Lightning-AI/lightning-thunder.git@main
 
 To achieve the best performance, you can install Thunder with the following additional dependencies:
 
-- install nightly [nvFuser](https://github.com/NVIDIA/Fuser) built for PyTorch 2.4 as follows:
+- install prerelease [nvFuser](https://github.com/NVIDIA/Fuser) built for PyTorch 2.5.1 as follows:
 
 ```bash
 # install nvFuser built for the matching stable PyTorch
-pip install --pre nvfuser-cu121-torch24
+pip install --pre nvfuser-cu121-torch25
 ```
 
 - install [cudnn](https://gitlab-master.nvidia.com/cudnn/cudnn_frontend) as follows:

docs/source/fundamentals/installation.rst

@@ -6,12 +6,12 @@ Minimal dependencies
 
 Follow these instructions to install PyTorch, nvFuser, and finally Thunder.
 
-Install PyTorch and nvFuser with pip (command shown is for CUDA 12.1 and PyTorch 2.4.x)::
+Install PyTorch and nvFuser with pip (command shown is for CUDA 12.1 and PyTorch 2.5.x)::
 
-  pip install --pre nvfuser-cu121-torch24
+  pip install --pre nvfuser-cu121-torch25
 
 cu121 can be replaced with cu118 depending on your CUDA version. NVFuser builds typically support the latest point release of PyTorch stable versions.
-For torch 2.4, cu124 is also supported. For nightly versions and more detailed instructions, please see https://github.com/NVIDIA/Fuser/#installation
+For torch 2.5, cu124 is also supported. For nightly versions and more detailed instructions, please see https://github.com/NVIDIA/Fuser/#installation
 
 You're all set with minimal dependencies, so you can follow `Install Thunder`_.