Get the app
fine-tune 0 0 0

polymer/dolphin-only-gpt-4

published by @polymer

Alignment dataset tracked in the /datasets sub-catalog.

Lifetime downloads
1.1K
Last 30 days
25
Likes
2
Size
1.4 GB
Created on HF
2023-09-06
Age
3 years ago

Description

Dolphin 🐬

https://erichartford.com/dolphin

Dataset details

This dataset is an attempt to replicate the results of Microsoft's Orca

Our dataset consists of:

~1 million of FLANv2 augmented with GPT-4 completions (flan1m-alpaca-uncensored.jsonl)

~3.5 million of FLANv2 augmented with GPT-3.5 completions (flan5m-alpaca-uncensored.jsonl)

We followed the submix and system prompt distribution outlined in the Orca paper. With a few exceptions. We included all 75k of CoT in the FLAN-1m… See the full description on the dataset page: https://huggingface.co/datasets/polymer/dolphin-only-gpt-4.

Tags

task_categories:text-generationlicense:apache-2.0size_categories:100K<n<1Mformat:jsonmodality:textlibrary:datasetslibrary:pandaslibrary:mlcroissantlibrary:polarsregion:us

README current version from Hugging Face

license: apache-2.0
task_categories:

  • text-generation
    duplicated_from: ehartford/dolphin

Dolphin 🐬

https://erichartford.com/dolphin

Dataset details

This dataset is an attempt to replicate the results of Microsoft's Orca

Our dataset consists of:

  • ~1 million of FLANv2 augmented with GPT-4 completions (flan1m-alpaca-uncensored.jsonl)
  • ~3.5 million of FLANv2 augmented with GPT-3.5 completions (flan5m-alpaca-uncensored.jsonl)

We followed the submix and system prompt distribution outlined in the Orca paper. With a few exceptions. We included all 75k of CoT in the FLAN-1m dataset rather than sampling that. Also, we found that many items were duplicated, so we removed duplicates, resulting in 3.5m instructs in the ChatGPT dataset.

Then we filtered out instances of alignment, refusal, avoidance, and bias, in order to produce an uncensored model upon which can be layered your personalized alignment LoRA.

Token distribution for GPT-3.5 completions
dolphin-llama

Loading

## load GPT-4 completions
dataset = load_dataset("ehartford/dolphin",data_files="flan1m-alpaca-uncensored.jsonl")

## load GPT-3.5 completions
dataset = load_dataset("ehartford/dolphin",data_files="flan5m-alpaca-uncensored.jsonl")

This dataset is licensed apache-2.0 for commercial or non-commercial use.

We currently plan to release Dolphin on:

  • Xgen 7b 8k
  • LLaMA 13b (Non-commercial)
  • MPT 30b 8k
  • LLaMA 33b (Non-commercial)
  • Falcon 40b
  • LLaMA 65b (Non-commercial)

The Dolphin models that are released will be subject to the license of the foundational model on which it is trained. (LLaMA releases will be non-commercial)

I would like to thank the motley crew of Open Source AI/ML engineers who have worked beside me in this endeavor. Including:

  • Wing "Caseus" Lian and NanoBit of OpenAccess AI Collective
  • Rohan
  • Teknium
  • Pankaj Mathur
  • Tom "TheBloke" Jobbins for quantizing and amplifying
  • Special thanks to EdenCoder and chirper.ai for mentorship and financial sponsorship.
  • Special thanks to Kilkonie for his very valued mentorship.
  • All the other people in the Open Source AI community who have taught me and helped me along the way.

README history 10 revisions

Every night we snapshot the README of every dataset in the catalog. When the SHA changes we archive the new version and diff it against the last. This is the evolving thought record of the alignment-data field: what the author decided to say about the corpus, and how that framing shifted over time.

  1. 2023-09-06Delete flan1m-alpaca-uncensored.jsonla92ca892.3 KB
    Loading...
  2. 2023-09-06Delete flan5m-alpaca-uncensored-deduped.jsonla6a9a662.3 KB
    Loading...
  3. 2023-09-06Delete flan5m-alpaca-uncensored.jsonled8dc242.3 KB
    Loading...
  4. 2023-09-06Delete flan5m-sharegpt-deduped.json994da182.3 KB
    Loading...
  5. 2023-09-06Delete flan1m-sharegpt-deduped.jsonbb55cc72.3 KB
    Loading...
  6. 2023-09-06Delete dedupeToShareGpt.pyd0baa772.3 KB
    Loading...
  7. 2023-09-06Delete convertToShareGpt.pyba8fbfa2.3 KB
    Loading...
  8. 2023-09-06Delete fp32_to_fp16.pyb4942bb2.3 KB
    Loading...
  9. 2023-09-06Delete llama_flash_attn_monkey_patch.py190b2bc2.3 KB
    Loading...
  10. 2023-09-06Duplicate from ehartford/dolphin047afa92.3 KB
    Loading...

Used by abliterated models none yet

No abliterated models in our catalog have declared this dataset in their YAML metadata yet. This can mean the dataset is used but not declared, is used outside abliteration workflows, or is new. The link will populate automatically as models are indexed.

Related datasets same stage · fine-tune

QuixiAI/dolphin
40.4K dl 444 likes 17.6 GB
Dolphin 🐬 https://erichartford.com/dolphin Dataset details This dataset is an attempt to replicate the results of ...
nvidia/Nemotron-SFT-Safety-v1
27.6K dl 14 likes 147.9 MB
Dataset Description: The Nemotron-SFT-Safety-v1 data is designed to align models to be robust against a variety of safety and s...
open-llm-leaderboard-old/details_ehartford__WizardLM-1.0-Uncensored-Llama2-13b
18.2K dl 293.9 MB
Dataset Card for Evaluation run of ehartford/WizardLM-1.0-Uncensored-Llama2-13b Dataset Summary Dataset automati...
jtatman/stable-diffusion-prompts-stats-full-uncensored
16.2K dl 157 likes 206.8 MB
jondurbin/airoboros-3.2
14.7K dl 52 likes 74.8 MB
Overview This dataset is a continuation of the airoboros-3.1 with the following changes: MathJSON has been removed for the tim...
xzuyn/open-instruct-uncensored-alpaca
12.2K dl 7 likes 7.4 GB
Original dataset page from ehartford. 810,102 entries. Sourced from open-instruct-uncensored.jsonl. Converted the jsonl to a json ...
OEvortex/uncensored-vortex
10.9K dl 10 likes 11.4 GB
jiae1234/vscd
10.5K dl 98.5 GB
VSCD: Video-based Scene Change Detection in Unaligned Scenes Dataset Description VSCD is a benchmark for Video-b...

Read further

Community best practices (data hygiene)
Rules for what belongs in a training mix and what does not.
M6 - Uncensored fine-tuning
DPO/ORPO from scratch. The Hartford dolphin lineage.
The politics of dataset curation
The curatorial choices behind Dolphin, Wizard, and the uncensored tradition.
Non-English abliteration communities
How non-English producers assemble uncensored fine-tuning corpora.

Metadata

License
apache-2.0

Repository

FILE .gitattributes FILE flan1m-alpaca-uncensored-deduped.jsonl FILE README.md

File contents live on Hugging Face. Click a file or folder to open it there.

Download the dataset

0

Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.

Recommended: datasets library (Python) 
from datasets import load_dataset
ds = load_dataset("polymer/dolphin-only-gpt-4")
Raw snapshot (Python) 
from huggingface_hub import snapshot_download
snapshot_download(repo_id="polymer/dolphin-only-gpt-4", repo_type="dataset")
git clone (requires git-lfs) 
git clone https://huggingface.co/datasets/polymer/dolphin-only-gpt-4
Browse repository on Hugging Face ->
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Get the app Browse the catalog
Open in Abliteration