Improve pte loading when given manual params_path

[Jack-Khuu]

Currently, when loading a .pte file is provided, the base model is loaded in order to obtain the model.config: ModelArgs. This is overkill (and is not resolved in this PR) since the rest of the base model is discarded.

This PR avoids this wasteful loading in one condition: a manual params_path is provided, which can be used to construct ModelArgs

---

Tested by passing in a pte w/ and without a manual param

python torchchat.py generate llama3.1 --pte-path model.pte
- Load

python torchchat.py generate llama3.1 --pte-path model.pte --params-path config.json
- No Load

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchchat/1178

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit fe90ac7 with merge base 72d2d20 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.