Rename Swift 1.5 fast variant to INT4 heads

Clarify model comparison and report English/Python perplexity

Clarify model comparison and report English/Python perplexity

Identify exact upstream checkpoint and revision before quantization changes

Remove introductory explanation paragraph from model card

Remove inherited packed-storage parameter-count override; preserve all weights

Lead with Swift plus HyperQwen task-time results and concise conversion details

Publish Swift HyperQwen collection with performance and quality comparisons

Clarify INT8-head baseline and fast quantization changes

Release Swift 1.5 HyperQwen fast quantization with RTX 3090 evaluation