model: fix all granite family parameter counts

[taronaeo]

Overview

cont: #28632

After #28632, I realised that quite a lot of our Granite family parameter counts are either inaccurate (2B model reported as 3B), missing, or using dense model naming schemes when the model is MoE or hybrid.

This PR fixes all Granite family parameter counts reported in llama-bench.

Additional information

-hf	Original	Updated
bartowski/granite-3.0-2b-instruct-GGUF:Q2_K	granite 2B Q2_K - Medium	No changes
bartowski/granite-3.0-8b-instruct-GGUF:Q2_K	granite 8B Q2_K - Medium	No changes
bartowski/granite-3.0-1b-a400m-instruct-GGUF:Q3_K_S	granitemoe 1B.A400M Q3_K - Small	No changes
bartowski/granite-3.0-3b-a800m-instruct-GGUF:Q2_K	granitemoe 3B Q2_K - Medium	granitemoe 3B.A800M Q2_K - Medium
bartowski/granite-3.1-2b-instruct-GGUF:Q2_K	granite 2B Q2_K - Medium	No changes
bartowski/granite-3.1-8b-instruct-GGUF:Q2_K	granite 8B Q2_K - Medium	No changes
bartowski/granite-3.1-1b-a400m-instruct-GGUF:Q2_K	granitemoe 1B.A400M Q2_K - Medium	No changes
bartowski/granite-3.1-3b-a800m-instruct-GGUF:Q2_K	granitemoe 3B Q2_K - Medium	granitemoe 3B.A800M Q2_K - Medium
bartowski/ibm-granite_granite-3.2-2b-instruct-GGUF:Q2_K	granite 2B Q2_K - Medium	No changes
bartowski/ibm-granite_granite-3.2-8b-instruct-GGUF:Q2_K	granite 8B Q2_K - Medium	No changes
ibm-granite/granite-3.3-2b-instruct-GGUF:Q2_K	granite 2B Q2_K - Medium	No changes
ibm-granite/granite-3.3-8b-instruct-GGUF:Q2_K	granite 8B Q2_K - Medium	No changes
ibm-granite/granite-4.0-micro-GGUF:Q2_K	granite 3B Q2_K - Medium	No changes
ibm-granite/granite-4.0-h-micro-GGUF:Q2_K	granitehybrid 3B Q2_K - Medium	No changes
ibm-granite/granite-4.0-h-tiny-GGUF:Q2_K	granitehybrid 7B.A1B Q2_K - Medium	No changes
ibm-granite/granite-4.0-h-small-GGUF:Q2_K	granitehybrid 32B Q2_K - Medium	granitehybrid 32B.A9B Q2_K - Medium
unsloth/granite-4.0-h-350m-GGUF:Q3_K_S	granitehybrid 350M Q3_K - Small	No changes
unsloth/granite-4.0-h-1b-GGUF:Q2_K	granitehybrid 1B Q2_K - Medium	No changes
ibm-granite/granite-4.1-3b-GGUF:Q2_K	granite 2B Q2_K - Medium	granite 3B Q2_K - Medium
ibm-granite/granite-4.1-8b-GGUF:Q2_K	granite 8B Q2_K - Medium	No changes
ibm-granite/granite-4.1-30b-GGUF:Q2_K	granite ?B Q2_K - Medium	granite 30B Q2_K - Medium
ibm-granite/granite-4.2-3b-GGUF:Q2_K	granite 2B Q2_K - Medium	granite 3B Q2_K - Medium
ibm-granite/granite-4.2-8b-GGUF:Q2_K	granite 8B Q2_K - Medium	No changes
ibm-granite/granite-4.2-30b-GGUF:Q2_K	granite ?B Q2_K - Medium	granite 30B Q2_K - Medium

cc: @gabe-l-hart

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: No, painstakingly checked every model 🙃

[gabe-l-hart]

Thank you for cleaning all this up! This has been on my TODO list for a while, but never come to the top. Your question about GraniteMoe w/ 40 layers surfaced that the source of that line was an accidental copy over during refactors, so it would be great to strip it out now while this cleanup is going on.

[taronaeo]

Latest push addressed review comments + removed the unused Granite MoE 40 layers check.

[gabe-l-hart]

Thanks!