Generate the card from the recipe: declare the quantization instead of hand-writing it

split_model.json

@@ -4,13 +4,23 @@
     "void_pass1",
     "void_pass2"
   ],
+  "quantized": true,
+  "quantization_bits": 4,
+  "quantization_group_size": 64,
   "recipe": "void-model",
   "source": "netflix/void-model",
   "license": "apache-2.0",
+  "quantization_scope": "transformer Linear weights only",
   "links": [
     "void-model-mlx (inference): https://github.com/dgrauet/void-model-mlx",
-    "VideoX-Fun-mlx (engine): https://github.com/dgrauet/VideoX-Fun-mlx",
-    "Base model weights: https://huggingface.co/dgrauet/CogVideoX-Fun-V1.5-5b-InP-mlx"
+    "VideoX-Fun-mlx (engine): https://github.com/dgrauet/VideoX-Fun-mlx"
   ],
-  "usage_url": "https://github.com/dgrauet/void-model-mlx"
+  "usage_url": "https://github.com/dgrauet/void-model-mlx",
+  "extra_links": [
+    "Base model weights (q8): https://huggingface.co/dgrauet/CogVideoX-Fun-V1.5-5b-InP-mlx-q8",
+    "bf16 variant: https://huggingface.co/dgrauet/void-model-mlx",
+    "q8 variant: https://huggingface.co/dgrauet/void-model-mlx-q8"
+  ],
+  "cli_snippet": "python -m void_mlx.infer \\\n    --sample sample/BigBen \\\n    --pass1 weights/q4/void_pass1.safetensors \\\n    --pass2 weights/q4/void_pass2.safetensors \\\n    --base-model /path/to/CogVideoX-Fun-V1.5-5b-InP-mlx-q8 \\\n    --steps 30 --max-frames 13 --height 352 --width 624 \\\n    --output result.gif",
+  "notes": "**This is the 32 GB configuration**: paired with the q8 base model, a full\ntwo-pass BigBen run (30 steps, 13 frames, 352\u00d7624) peaks at ~23.7 GB \u2014\nunder the 26.8 GB recommended working set of a 32 GB Apple Silicon Mac.\nQuality vs the bf16 weights: PSNR \u2248 35.5 dB on the same seed."
 }