dolphin-8b-merged
Standalone merged model: dphn/dolphin-2.9-llama3-8b + Cutyp/dolphin-8b-qlora.
The QLoRA adapter was merged into the base weights with PEFT scaling
alpha / r = 32 / 16 = 2.0, so this repo runs without any adapter code:
from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("Cutyp/dolphin-8b-merged")
model = AutoModelForCausalLM.from_pretrained("Cutyp/dolphin-8b-merged")
- Base:
dphn/dolphin-2.9-llama3-8b - Adapter:
Cutyp/dolphin-8b-qlora(r=16, alpha=32, targets q/k/v/o/gate/up/down_proj) - Merged: 2026-10-01, shard-by-shard float32 merge, original dtypes preserved
- Intended repo name:
Cutyp/dolphin-8b-merged