grimjim
/

kukulemon-32K-7B

Text Generation

text-generation-inference

Inference Endpoints

Model card Files Files and versions Community

grimjim commited on Mar 30

Commit

a4d40e8

•

1 Parent(s): 941fd3f

Update README.md

Updated with comparison of SLERP versus DARE-TIES

Files changed (1) hide show

README.md +2 -1

README.md CHANGED Viewed

@@ -11,8 +11,9 @@ license: cc-by-nc-4.0
 ---
 # kukulemon-32K-7B
-This is an attempt at a merge capable of functional 32K context length while being derived from [kukulemon-7B](https://huggingface.co/grimjim/kukulemon-7B).
 The functioning 32K context window has been folded in via a merger of Mistral 7B v0.2 models.
 Although the resulting model natively supports Alpaca prompt, I've tested with ChatML prompts successfuly. Medium temperature (around 1) with low minP (e.g., 0.01) works with ChatML prompts in my most recent testing.

 ---
 # kukulemon-32K-7B
+This is a proof of concept a merge capable of functional 32K context length while being derived from [kukulemon-7B](https://huggingface.co/grimjim/kukulemon-7B).
 The functioning 32K context window has been folded in via a merger of Mistral 7B v0.2 models.
+SLERP merge appears to be viable, but DARE-TIES merge risks producing a damaged model and is therefore not recommended.
 Although the resulting model natively supports Alpaca prompt, I've tested with ChatML prompts successfuly. Medium temperature (around 1) with low minP (e.g., 0.01) works with ChatML prompts in my most recent testing.