Repoint build instructions to charlie12345/ROCmFPX (ROCmFPX FP3/4/6/8 repo)

Upload README.md with huggingface_hub

Upload README.md with huggingface_hub

consolidate to single best-balance rocmfp4 quant

consolidate to single best-balance rocmfp4 quant

Card: best-of-both recommended + KL findings + bundled mmproj/vision flags

Add mmproj-F32.gguf (best-of-both / vision)

Add Qwen3.6-27B-MTP-ROCmFP4-COHERENT-imatrix-embF16-headQ6.gguf (best-of-both / vision)

Center top bar; remove table bottom-margin gap

Center logo+title pair, structured subtext, full-width spec table (display:table)

Shrink logo to 5px

Shrink logo to 6px, center it when stacked on mobile

Move Qwen logo inline left of title (responsive, borderless, smaller)

Add Qwen ASCII logo banner to card header

Correct mmproj/cache claim: source says only --cache-reuse is disabled w/ multimodal (not 'cache reuse entirely'); rest unmeasured. (27B: + upstream-benchmark pointer)

Card: datasheet/neo-brutalist restyle (monospace, hard borders, spec grid, numbered sections); claims audited (f16 KV stated as config only)

Card: add Vision section (--mmproj + --image-min-tokens 1024) and bundled chat_template.jinja note

Add froggeric v20 fixed chat template (vision + inline think-toggle + tool anti-stall); use via --chat-template-file

provenance: set base_model_relation=quantized

Card: warn that HF's auto badge (F16) is wrong; these are 4-bit ROCmFP4

Card: document Q6-head variant (measurement + cost/benefit)

Add Q6-head variant (output.weight -> Q6_K): more faithful, ~5-7% slower decode

Upload README.md with huggingface_hub

Upload README.md with huggingface_hub

Upload README.md with huggingface_hub

Upload README.md with huggingface_hub

Upload Qwen3.6-27B-MTP-ROCmFP4-STRIX-imatrix-embF16.gguf with huggingface_hub

Upload qwen3.6-27b-code.imatrix with huggingface_hub

Upload README.md with huggingface_hub

initial commit