olmoe HF conversion

[sohamparikh]

✨ Description

OLMoE HF conversion support in FastLLM.

• norm_topk_prob is not supported in FastLLM and hence hardcoded
• qk_norm=False is only supported in this fork but not in the upstream transformers repo. Upstream transformers will apply RMSNorm to query and key projections by default

Considering these, the current Olmoe implementation in Fast-LLM is no different than that of Mixtral, and Mixtral's converter should work as well.

Partially resolves #61

🔍 Type of change

Select all that apply:

• 🐛 Bug fix (non-breaking change that addresses a specific issue)
• 🚀 New feature (non-breaking change that adds functionality)
• ⚠️ Breaking change (a change that could affect existing functionality)
• 📈 Performance improvement/optimization (improves speed, memory usage, or efficiency)
• 🛠️ Code refactor (non-functional changes that improve code readability, structure, etc.)
• 📦 Dependency bump (updates dependencies, including Dockerfile or package changes)
• 📝 Documentation change (updates documentation, including new content or typo fixes)
• 🔧 Infrastructure/Build change (affects build process, CI/CD, or dependencies)

📝 Changes

List the key changes introduced in this PR:

1. HF Converter for OLMoE

✅ Checklist

Make sure the following tasks are completed before submitting the PR:

General

• 📜 I have read and followed the contributing guidelines.
• 🏷️ I am using a clear and descriptive PR title that summarizes the key change or feature introduced.
• 🎉 The functionality is complete, and I have tested the changes.
• 📝 I have updated the documentation if needed.
• ⚠️ The change does not introduce any new issues (e.g., runtime warnings, type checker errors, linting problems, unhandled edge cases).
• 🧩 I have commented my code, especially in hard-to-understand areas.

Dependencies and Configuration

• 🐋 I have updated the Docker configuration or dependencies, if applicable.
• 🔄 I have ensured compatibility with the existing setup after dependency changes.

Testing

• 🧪 I have added or updated tests to cover my changes.
• ✔️ New and existing tests pass locally with my changes.
• 🚦 I have tested these changes on GPUs and verified training stability.
• 🏋️ I have tested the changes on realistic training workloads, if applicable.

Performance Impact

• 📊 I have run benchmarks where applicable to evaluate the performance impact.
• ✅ The benchmarks show no performance regression.
• 🚀 The benchmarks indicate a potential performance improvement.
• ⚠️ The benchmarks indicate a potential performance degradation.
• 📈 I have provided benchmark results and detailed any performance impact below, if applicable.

📊 Performance Impact Details

If there is any impact on performance, describe it and provide benchmark results, if applicable:

---

🗒️ Additional Notes

Include any additional context, information, or considerations here, such as known issues, follow-up tasks, or backward compatibility concerns.

[jlamypoirier]

Are we planning to merge this? If so:

• Fix llama conversion, improve parameter conversion #94 changed config conversion a bit, so some changes are needed in _create_config_converters. See Fix llama conversion, improve parameter conversion #94 for examples.
• Consider adding a olmoe model to the test suite (tests/common.py). It can just be a copy of mixtral with a different converter if that works. Then you can test the converter with MODEL=olmoe pytest tests/test_checkpoint.py.

[tscholak]

We have been using the Mixtral exporter, I believe we don't need this converter anymore.