[codex] fix LFM2 GGUF conversion fallback

[i386]

This fixes GGUF conversion for the original LiquidAI/LFM2-350M checkpoint when convert_hf_to_gguf.py cannot find LFM2 FFN metadata in self.hparams.

What changed

• add a small LFM2-specific helper that falls back to raw config.json values for:
  • block_ff_dim
  • block_auto_adjust_ff_dim
  • block_ffn_dim_multiplier
  • block_multiple_of
• keep explicit KeyErrors if those values are truly missing

Why

Converting the original LiquidAI/LFM2-350M checkpoint currently fails with a KeyError on block_ff_dim even though the value is present in config.json.

This blocks same-origin GGUF conversion for LFM2 text checkpoints.

Impact

Users can convert original LFM2 text checkpoints to GGUF without patching the converter locally.

Validation

• converted LiquidAI/LFM2-350M to GGUF f16
• quantized the result to Q4_K_M
• validated the converted artifact locally against our exact smoke suite

[ggml-gh-bot]

Hi @i386, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• AI-generated content: This project does not accept PRs, descriptions or commit messages that are fully or predominantly AI-generated. If you have used AI to assist you in writing code, please make sure to disclose that explicitly.

• Large PR: Large changes require prior discussion (e.g. an issue or RFC) and maintainers may not be able to review this PR as-is. Consider splitting it into smaller, focused PRs.

---

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.

[CISC]

This, while working, is not entirely accurate, see #21508 for the appropriate fix.

[i386]

Ill close this then @CISC 👍 - thank you