Update README.md

README.md

@@ -16,14 +16,17 @@ This is a Llama 2-based model consisting of a gradient merge between:
 - [Chronos 13b v2](https://huggingface.co/elinas/chronos-13b-v2)
 - [Nous Hermes Llama2 13b](https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b)
 
+Quantized Models Provided by TheBloke (Thanks!):
+- [GGML](https://huggingface.co/TheBloke/Chronohermes-Grad-L2-13B-GGML)
+- [GPTQ](https://huggingface.co/TheBloke/Chronohermes-Grad-L2-13B-GPTQ)
+
+
 The merge was performed using [BlockMerge_Gradient](https://github.com/Gryphe/BlockMerge_Gradient) by Gryphe
 
 The intended objective was to combine NH2's superior instruction following capabilities with the creativity and response length of Chronos v2. Merge ratios used are identical to those used in [Chronoboros Grad](https://huggingface.co/kingbri/chronoboros-grad-l2-13B), with NH2 starting with a weight of 0.9 at the 1st layer and phasing out by the 25th layer. The method is illustrated in the image below, with green representing NH2 and blue representing Chronos v2:
 
 ![hermeboros-illustration](https://files.catbox.moe/18sjej.png)
 
-added_tokens.json was padded with dummy tokens to reach 32 added tokens in order to allow GGML conversion in llama.cpp without error due to vocab size mismatch.
-
 ## Usage:
 
 Intended to be prompted with the Alpaca instruction format of the base models: