Add Orpheus full fine-tune weights + base tokenizer

5b2d586 verified 4 months ago

992 Bytes

license: apache-2.0
language:
  - en
tags:
  - text-to-speech
  - orpheus
  - llama
base_model: canopylabs/orpheus-3b-0.1-pretrained

Orpheus full fine-tune (merged weights)

This repository contains full fine-tuned causal LM weights from vaghawan/eng-pigdin-mix-full-ft-mar-29-orpheus (checkpoint-1408), packaged with the tokenizer from canopylabs/orpheus-3b-0.1-pretrained for inference with the Orpheus / SNAC stack.

Load

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("vaghawan/orpheus-model-full", torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("vaghawan/orpheus-model-full")

Source

Weights: vaghawan/eng-pigdin-mix-full-ft-mar-29-orpheus checkpoint-1408
Tokenizer: canopylabs/orpheus-3b-0.1-pretrained