Support llama3.1-8b-Instruct in to_huggingface

[hengtaoguo]

Description

Complete llama3.1-8b-Instruct to_huggingface conversion, by using the base model's mapping information.

Static Assets: Specialized assets (the Instruct tokenizer, tokenizer.json, and templates) continue to be downloaded and saved correctly via tokenizer.save_pretrained using the -Instruct Hub ID.

Fixes: b/510847180

Tests

Checkpoint conversion:

python -m maxtext.checkpoint_conversion.to_huggingface model_name=llama3.1-8b-Instruct load_parameters_path=gs://hengtaoguo-maxtext-logs/checkpoints/llama3.1-8b-Instruct/scanned/2026-05-08/0/items base_output_directory=/dev/shm/hengtaoguo/hf_safetensor/llama3.1-8b-Instruct scan_layers=true weight_dtype=bfloat16 hardware=cpu skip_jax_distributed_system=True hf_access_token=<your_token>

Forward pass logit check:

python -m tests.utils.forward_pass_logit_checker maxtext/configs/base.yml run_name=ht_test model_name=llama3.1-8b tokenizer_path=meta-llama/Llama-3.1-8B-Instruct load_parameters_path=gs://hengtaoguo-maxtext-logs/checkpoints/llama3.1-8b-Instruct/scanned/2026-05-08/0/items max_prefill_predict_length=4 hf_access_token=<your_token> --run_hf_model=True --hf_model_path=meta-llama/Llama-3.1-8B-Instruct --max_kl_div=0.015

Logs

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!