Parameter to optionally override default tokenizer path

[coleramos425]

I've found that when running llm.c out of directory, following the instructions provided in README:

pip install -r requirements.txt
python dev/data/tinyshakespeare.py
python train_gpt2.py
make train_gpt2amd
./train_gpt2amd

I get the following error:

---
WARNING: Failed to open the tokenizer file gpt2_tokenizer.bin
The Tokenizer is a new feature added April 14 2024.
Re-run `python train_gpt2.py` to write it
---

I needed to add a parameter to the train_gpt2.cu file to allow user to specify path to tokenizer file. Running with this change fixed the issue for me.