Sync model repo (text/metadata)

benchmarks/README.md

@@ -1,7 +1,7 @@
 # Benchmarks
 
 Machine-readable measurement records for this model, conforming to
-[`mech-model-schemas`](https://github.com/Arm-Debug/mech-model-schemas)
+the AI Portal model schema
 v1.0.0 (`task: vision-classification`).
 
 These complement — they don't replace — the eval block in the model
@@ -48,7 +48,7 @@ the disambiguator is `weight_dtype` (and, once published under an
 
 ## Provenance
 
-Source: Marvik delivery package `swin_tiny_raspberry_executorch.zip`
+Source: the delivery package `swin_tiny_raspberry_executorch.zip`
 (`report.json`, `metadata.yaml` / `metadata_raspberry.yaml`,
 `huggingface/README.md`). The records map from `report.json` (real Pi 5
 hardware run); the FP32 baseline is reconstructed from its `original`

benchmarks/swin-tiny-raspberrypi5-fp32.yaml

@@ -2,7 +2,7 @@
 # Pairs with `swin-tiny-raspberrypi5-int8.yaml` to demonstrate the
 # baseline-comparison join.
 #
-# Source: Marvik delivery package `swin_tiny_raspberry_executorch.zip`,
+# Source: the producer delivery package `swin_tiny_raspberry_executorch.zip`,
 # `report.json` `original` fields (the un-quantised baseline measured under
 # identical target / runtime / dataset / benchmark conditions to the optimised
 # record).
@@ -10,7 +10,7 @@
 # Same target / runtime / dataset / benchmark as the INT8 record — the
 # disambiguator is `weight_dtype` (and, once published, `model.id`).
 #
-# Hand-curated rather than redelivered: Marvik's pipeline only emits the
+# Hand-curated rather than redelivered: the producer's pipeline only emits the
 # optimised metadata.yaml, so this baseline is reconstructed from report.json.
 #
 # Note: the delivery's vendored huggingface/metadata.yaml is the *Emulation*