Upload README.md with huggingface_hub

README.md

@@ -51,6 +51,12 @@ base_model = AutoModelForCausalLM.from_pretrained(base, trust_remote_code=True,
 model = PeftModel.from_pretrained(base_model, adapter)
 ```
 
+## Training loss vs training progress
+
+![Training loss vs progress](./training_loss_vs_progress.svg)
+
+Loss curve included as optimization check; qualitative rewrite fidelity remains the decision metric.
+
 ## Recommended evaluation sample
 Use this full fiction passage for held-out testing: