Luna-AI-Llama2-Uncensored
Llamauncensored
from Tap-M/Luna-AI-Llama2-Uncensored
↓ 10.2K♥ 14614 repos
dolphin-llama3-zh-cn-uncensored
dolphin-llama3-zh-cn-uncensored is a abliterated and uncensored variant of dphn/dolphin-2.9-llama3-8b, published on Hugging Face by stephenlzc. It is mit licence. It is distributed across 2 repositories in GGUF format, totalling 5.7K downloads.
Specification
- Model name
- dolphin-llama3-zh-cn-uncensored
- Source repository
- stephenlzc/dolphin-llama3-zh-cn-uncensored
- Base model
- dphn/dolphin-2.9-llama3-8b
- Publisher
- stephenlzc
- Parameters
- Not determinable from name
- Hardware class
- Unknown
- Licence
- mit
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF
- First indexed
- 21 Jun 2024
- Last updated
- 24 Jun 2024
Ablation technique
How dolphin-llama3-zh-cn-uncensored was modified, and what that implies.
Abliteration + uncensored fine-tune
Directional ablation combined with additional fine-tuning on unfiltered data, so behaviour diverges from the base model beyond refusal removal alone.
Reported scope: Not stated by the publisher. Publishers frequently omit this, so verify behaviour empirically rather than assuming full coverage.
From the publisher’s model card
“Using cognitivecomputations/dolphin-2.9-llama3-8b as base model, and finetune with Minami-su/toxic-sft-zh and llm-wizard/alpaca-gpt4-data-zh to let tha model support Chinese.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download stephenlzc/dolphin-llama3-zh-cn-uncensored --local-dir ./dolphin-llama3-zh-cn-uncensored
llama-cli -m ./dolphin-llama3-zh-cn-uncensored/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "stephenlzc/dolphin-llama3-zh-cn-uncensored"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve stephenlzc/dolphin-llama3-zh-cn-uncensored --trust-remote-codeDownloads and variants (1)
The only published repository of dolphin-llama3-zh-cn-uncensored.
| Repository | Format | Downloads |
|---|---|---|
| stephenlzc/dolphin-llama3-zh-cn-uncensoredsource | GGUF | 5.7K |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- dolphin-llama3-zh-cn-uncensored will attempt offensive-security prompts that a hosted commercial model declines, which is what makes it usable for red-team corpus generation and for measuring what an unaligned model of this class produces.
- The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
- This release adds fine-tuning on unfiltered data, so its behaviour diverges from the base model beyond refusal removal — differences you measure cannot be attributed to ablation alone.
- Declared licence is mit, but the terms that bind you are dphn/dolphin-2.9-llama3-8b’s — a re-release cannot grant more than its parent.
Handling: run it isolated, put moderation in front of it if anyone outside your team can reach it, and verify the weights before loading — these are third-party artifacts. Full handling and licence guidance.
Metadata for dolphin-llama3-zh-cn-uncensored is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with stephenlzc, and does not independently verify publisher claims. See responsible use.