Hugging Face's logo

Quant-Cartel
/
Dusk-Miqu-70B-iMat-GGUF

GGUF
GGUF
iMat
Inference Endpoints
imatrix
Model card Files Community
InferenceIllusionist commited on
Commit
3d5fe52
•
1 Parent(s): 8016a2c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -0
README.md CHANGED
@@ -32,6 +32,7 @@ PROUDLY PRESENTS
32
  Quantized from fp16.
33
  * Weighted quantizations were creating using fp16 GGUF and [groups_merged-enhancedV2-TurboMini.txt](https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-9432658) in 234 chunks and n_ctx=512
34
  * This method of calculating the importance matrix showed improvements in some areas for Mistral 7b and Llama3 8b models, see above post for details
 
35
 
36
  For a brief rundown of iMatrix quant performance please see this [PR](https://github.com/ggerganov/llama.cpp/pull/5747)
37
 
 
32
  Quantized from fp16.
33
  * Weighted quantizations were creating using fp16 GGUF and [groups_merged-enhancedV2-TurboMini.txt](https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-9432658) in 234 chunks and n_ctx=512
34
  * This method of calculating the importance matrix showed improvements in some areas for Mistral 7b and Llama3 8b models, see above post for details
35
+ * The enhancedv2-turbomini file appends snippets from turboderp's calibration data to the standard groups_merged.txt file
36
 
37
  For a brief rundown of iMatrix quant performance please see this [PR](https://github.com/ggerganov/llama.cpp/pull/5747)
38