[core] restrict logging of quant config based on the header size.

[sayakpaul]

Cc: @asomoza does it work?

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[asomoza]

thanks a lot @sayakpaul, yeah this fixes the issue I was having

[sayakpaul]

@DN6 could you review it? This PR unblocks the loading SDNQ checkpoints. Currently, we log the quant config (if user doesn't disable telemetry) in user_agent. We do this for granular comparison on how quant backends are used. For example, is FP8 more used in TorchAO compared to INT8?

However, SDNQ's quant configs can get quite big and can fail the user_agent. Another alternative I considered was the send_telemetry() function from huggingface_hub which is non-blocking. But decided to go with this approach.

[yiyixuxu]

thanks!