Fix Gpt3 MultiHeadAttention out projection dimension

[hsuan-lun-chiang]

Description

The output dimension of the output projection in the Gpt3 MultiHeadAttention layer is mistakenly set to self.num_heads * self.head_dim; it should be config.emb_dim.

Tests

python3 -m MaxText.train  MaxText/configs/base.yml run_name=gpt3-train-run base_output_directory=gs://maxtext-test/train_gpt3/13/ model_name=gpt3-6b dataset_type=synthetic steps=10

Logs: base_emb_dim=3584
Logs: base_emb_dim=3584 base_mlp_dim=3584

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.