Publish verified portfolio asset

README.md

@@ -0,0 +1,22 @@
+---
+library_name: torchvision
+pipeline_tag: image-classification
+tags:
+- efficientnet
+- aws
+- sagemaker
+- healthcare-ai
+- reference-card
+---
+# OtoSage EfficientNetV2-S Reference
+
+Architecture/provenance card for the OtoSage SageMaker training path.
+
+- `torchvision.models.efficientnet_v2_s`
+- pretrained dependency: `EfficientNet_V2_S_Weights.DEFAULT`
+- project code replaces the classifier for the otoscopic classes
+
+**No user-trained SageMaker checkpoint is hosted here.** The roadmap requires an executed SageMaker training/evaluation/model-registry run before a checkpoint or cloud-performance claim is published.
+
+Source revision: `eb2188bbf0b42751c35a53af20947bae0d3c994e`  
+Source: https://github.com/singhankitsrf/OtoSage_AWS_GitHub

model_spec.json

@@ -0,0 +1,10 @@
+{
+  "framework": "torchvision",
+  "architecture": "efficientnet_v2_s",
+  "pretrained_weights_reference": "EfficientNet_V2_S_Weights.DEFAULT",
+  "classifier_head": "Dropout(0.25 default) + Linear(in_features, num_classes)",
+  "source_file": "ml/src/train.py",
+  "source_revision": "eb2188bbf0b42751c35a53af20947bae0d3c994e",
+  "user_trained_checkpoint_included": false,
+  "deployment_target": "SageMaker asynchronous inference after validated training"
+}