Create README.md

README.md

@@ -0,0 +1,19 @@
+# Full finetuning
+
+This model finetunes the pretrained model `canopylabs/orpheus-3b-0.1-pretrained` using synthetic data generated by selective zeroshot voice cloning of higgs-audio. Full finetuning with Unsloth for 1 epoch.
+
+### Dataset
+`vietnhat/mapalo-higgs-metadata2-v2`
+`vietnhat/dee-higgs-metadata1-v2`
+`vietnhat/zenande-higgs-metadata3-v2`
+`vietnhat/tyra-higgs-metadata4-v2`
+
+### Inference
+
+Available voice: `dee`, `mapalo`, `tyra`, `zenande`
+
+```bash
+temperature = 0.7
+top_p = 0.9
+repetition_penalty = 1.1
+```