Remove internal maintainer-warning comment from model card

README.md

@@ -44,23 +44,6 @@ model-index:
 
 This is an INT8-quantized version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) optimized for edge deployment on ARM devices using [ExecuTorch](https://github.com/pytorch/executorch) with the XNNPACK backend. The model uses 8da8w dynamic quantization for the 192 encoder/decoder Transformer-stack Linear weights via Optimum-ExecuTorch, plus a manual weight-only INT8 quantization pass on `decoder.embed_tokens`, and is exported to the `.pte` format for efficient inference on ARM Cortex-A processors (AWS Graviton, mobile ARM, embedded). The final output projection (`proj_out` / `lm_head`), positional embeddings, and encoder convolutional stem remain FP32.
 
-<!--
-MAINTAINER WARNING: both Whisper Small model weights are NOT in git.
-
-The optimized INT8 `whisper-small-int8-executorch.pte` (395.0460 MB), FP32
-baseline `whisper-small-fp32-executorch.pte` (1074.6592 MB), and matching
-preprocessor `.pte` files are gitignored and kept only in local working copies
-until the large-weight publishing path is fixed. The model card, config,
-benchmarks, and sample audio are tracked.
-
-CRITICAL: `.github/workflows/publish.yml` uses `hf upload --delete '*'`. The
-workflow now fails fast for Whisper when these ignored files are missing, because
-a publish from git alone cannot upload the two model weights. If that guard is
-bypassed, manually uploaded copies can still be deleted from the HF repo. Re-upload
-both files by hand after any successful publish until the permanent large-file
-solution replaces this stopgap.
--->
-
 ## Key Highlights
 
 Compared to the FP32 baseline: