Onboard GPT OSS model

[RissyRan]

Description

Onboard GPT OSS model:

• Add two GPT OSS model configs
• Add necessary configs in base.yml file (e.g., mlp_activations_limit, mlp_bias, attention_bias, attention_sink)
• Add sink feature for dot_product and enable for flash_attention
• Add decoder layers for scan and unscan features

Next step: we will work on ckpt conversion and verify logits

Tests

• Unit test check_gpt_vs_reference.py against reference implementation for attention & mlp blocks
• End-to-end functional tests for scan and unscan on gpt-oss-20b model

Checklist

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

• I have performed a self-review of my code.
• 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.

[shuningjin]

Thank you for the great work!

[gagika]

A few more comments, if you agree, feel free to address them in a separate PR.

[aireenmei]

Thanks Ran!

[gagika]

Thanks