Replete-AI_-_Phi-Delthanar-gguf

Quantization made by Richard Erkhov.

Github

Discord

Request more models

Phi-Delthanar - GGUF

• Model creator: https://huggingface.co/Replete-AI/
• Original model: https://huggingface.co/Replete-AI/Phi-Delthanar/

Original model description:

license: mit language:

• en thumbnail: "https://cdn-uploads.huggingface.co/production/uploads/6589d7e6586088fd2784a12c/iYImJKf2HZZZJ9IwDSN00.png"

---

The forest is with you.

Named after the method used to create it, interleaving the layers of its predecessor to become far larger, giving it much more potential.

Del'thanar is a supposed ancient treeant, and I couldn't think of a better naming convention for a model that was created using the passthrough method.

By concatenating layers from different LLMs, it can produce models with an exotic number of parameters (e.g., 9B with two 7B parameter models). These models are often referred to as "frankenmerges" or "Frankenstein models" by the community.

Many thanks to Abacaj for providing the fine tuned weights that were used in the creation of this base model. You can find the full script for how the model was merged here...thanks to KatyTheCutie for inspiring me to test out this script.

This idea was brought to me by The Face of Goonery, also known as Caleb Morgan. I have him to thank if fine-tuning this model turns out to be a success...he also helped me to make this model even larger than the prior one.

How to run inference:

import transformers
import torch

if __name__ == "__main__":
  model_name = "Replete-AI/Phi-Delthanar"
  tokenizer = transformers.AutoTokenizer.from_pretrained(model_name)
  
  model = (
      transformers.AutoModelForCausalLM.from_pretrained(
          model_name,
      )
      .to("cuda:0")
      .eval()
  )
  
  messages = [
      {"role": "user", "content": "Hello, who are you?"}
  ]
  inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
  input_ids_cutoff = inputs.size(dim=1)
  
  with torch.no_grad():
      generated_ids = model.generate(
          input_ids=inputs,
          use_cache=True,
          max_new_tokens=512,
          temperature=0.2,
          top_p=0.95,
          do_sample=True,
          eos_token_id=tokenizer.eos_token_id,
          pad_token_id=tokenizer.pad_token_id,
      )
  
  completion = tokenizer.decode(
      generated_ids[0][input_ids_cutoff:],
      skip_special_tokens=True,
  )
  
  print(completion)

Chat template

The model uses the same chat template as found in Mistral instruct models:

Join the Replete AI Discord here!