Upload README.md with huggingface_hub

README.md

@@ -17,7 +17,7 @@ pipeline_tag: image-classification
 Pre-trained model weights from [vladmandic/face-api](https://github.com/vladmandic/face-api)
 at commit `189226d63aabb48cb40776fd1c453ebc0fa722f1`, converted losslessly to
 portable **safetensors** and **gguf** (float32) formats by the
-[`CognitiveOS-Labs/face-recognition`](https://github.com/CognitiveOS-Labs/face-recognition)
+[`CognitiveOS-Labs/tfjs-weights-to-gguf`](https://github.com/CognitiveOS-Labs/tfjs-weights-to-gguf)
 reproducible pipeline.
 
 ## Models
@@ -52,7 +52,8 @@ and the original TF.js quantization block (source traceability).
   load them yet — a runtime for these architectures is the documented
   follow-up (see the `face-recognition` CGP repo).
 - Reference loaders and the full bitwise-validation suite live in the
-  `face-recognition` CGP repo under `scripts/`.
+  [`CognitiveOS-Labs/tfjs-weights-to-gguf`](https://github.com/CognitiveOS-Labs/tfjs-weights-to-gguf)
+  repo under `scripts/`.
 
 ## Consumed by
 
@@ -66,7 +67,7 @@ and the original TF.js quantization block (source traceability).
 
 Weights derived from `vladmandic/face-api` (MIT), commit
 `189226d63aabb48cb40776fd1c453ebc0fa722f1`, provided "as-is". Conversion tooling
-in the `face-recognition` CGP repo is MIT.
+in the `tfjs-weights-to-gguf` repo is MIT.
 
 The recognition backbone was trained on VGGFace2-like data by the face-api
 project; generic face recognition works out of the box, while recognizing