Downloads · 30 days
15
6% of all-time downloads
cpral/Hermes-4-405B-exl3-5bpw
Hermes-4-405B-exl3-5bpw is a text generation model from cpral. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as llama3.
5.00 BPW H8 exllamav3 quant of Hermes 4 405B.
Downloads · 30 days
15
6% of all-time downloads
All-time downloads
256
Public
Parameters
129B
257 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors257 GB · 100%
How the weights are stored.
I16127B · 98%
From the Hugging Face model README
5.00 BPW H8 exllamav3 quant of Hermes 4 405B.
-- A perplexity: 1.50484401
-- B perplexity: 1.52952951
-- A label in top-K:
K = 1: 0.8938
K = 2: 0.9486
K = 3: 0.9640
K = 4: 0.9714
K = 5: 0.9757
-- B label in top-K:
K = 1: 0.8893
K = 2: 0.9466
K = 3: 0.9628
K = 4: 0.9702
K = 5: 0.9747
-- Top-K agreement, A vs B:
K = 1: 0.9722
K = 2: 0.8527
K = 3: 0.6881
K = 4: 0.5185
K = 5: 0.3663
-- KL divergence (A, B): 0.01720790
-- KL divergence (B, A): 0.01654647
command used to generate this quant
ulimit -n 100000
PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True uv run python convert.py -i /path-to/Hermes-4-405B \
-o /path-to/final/hermes4-405b-5bpw \
-w /path-to/workdir \
-b 5.0 \
-hq \
-ss 2048 \
-cpi 3600 \
-hb 8 \
-pm \
-d 0,1,2,3,4,5,6,7 \
-dr 1,1,1,1,1,1,1,1
<img src="https://cdn-uploads.huggingface.co/production/uploads/6317aade83d8d2fd903192d9/roT9o5bMYBtQziRMlaSDf.jpeg" width="300" style="float:center" />