|
|
--- |
|
|
base_model: |
|
|
- cognitivecomputations/dolphin-2.9.4-llama3.1-8b |
|
|
--- |
|
|
|
|
|
# GGUF quants of cognitivecomputations/dolphin-2.9.4-llama3.1-8b, here I have: |
|
|
``` |
|
|
IQ4_XS (4.2G, 8.7992 +/- 0.11237, fits into 8GiB VRAM + 4096 context with F16 KV cache) |
|
|
Q4_K_M (4.6G, 8.7948 +/- 0.11223, fits into 8GiB VRAM + 4096 context with F16 KV cache, also good for CPU inference on E5-26xx v3/v4) |
|
|
Q8_0 (8.0G, 8.5970 +/- 0.10933, imatrix derived from it) |
|
|
F16 ( 15G, 8.6617 +/- 0.11043, for 24GiB VRAM) |
|
|
``` |
|
|
Perplexity measured with `-fa -c 2048 -ub 2048` on UTF-8 text version of ["Wired Love" from Project Gutenberg](http://www.gutenberg.org/ebooks/24353). |
|
|
|
