LLM EXPLORER 58,297 MODELS INDEXED

Saily 220B by deepnight-research

By deepnight-research · 96 downloads
HF Repo ↗ Copy URL

Saily 220B is an open-source language model by deepnight-research. Features: 220b LLM, VRAM: 417GB, Context: 4K, License: llama2, LLM Explorer Score: 0.1.

  Dataset:eleutherai/pile   Dataset:meta-math/metamathqa Dataset:tiiuae/falcon-refinedw...   En   Endpoints compatible   Llama   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/deepnight-research/Saily_220B 

Saily 220B Parameters and Internals

Model Type 
AI model, text generation
Use Cases 
Limitations:
Saily_220B may generate incorrect or biased content.
Additional Notes 
Please don't refer to the config.json in the files, it isn't accurate.
Training Details 
Data Sources:
tiiuae/falcon-refinedweb, EleutherAI/pile, meta-math/MetaMathQA
Methodology:
The models were fine-tuned on a part of Refined-Web Dataset and individually the models were finetuned on niche specific datasets: Code, Humor, Maths, Logical Understanding, Physics, Reasoning, Psychology, Roleplay. The private dataset was created by transcribing internal brainstorming sessions and sessions with experts like mathematicians, developers, bio-engineers, authors, psychologists, etc.
Model Architecture:
Llama2 70B merges and fine-tuned models
Input Output 
Input Format:
Alpaca Prompt Format
Accepted Modalities:
text
Performance Tips:
Load the model in 4bit to fit on 2 x A100 (80GB).
Release Notes 
Version:
v1
Date:
17th December, 2023
Notes:
Saily_220B is a powerful AI model built on top of Llama2-70B merges. Includes fine-tuned and linearly merged models.
LLM NameSaily 220B
Repository 🤗https://huggingface.co/deepnight-research/Saily_220B 
Model Size220b
Required VRAM417 GB
Updated2026-08-04
Maintainerdeepnight-research
Model Typellama
Model Files  9.9 GB: 1-of-43   9.9 GB: 2-of-43   10.0 GB: 3-of-43   9.7 GB: 4-of-43   10.0 GB: 5-of-43   9.9 GB: 6-of-43   9.8 GB: 7-of-43   9.8 GB: 8-of-43   9.8 GB: 9-of-43   9.8 GB: 10-of-43   9.7 GB: 11-of-43   9.6 GB: 12-of-43   9.6 GB: 13-of-43   9.8 GB: 14-of-43   9.8 GB: 15-of-43   9.7 GB: 16-of-43   10.0 GB: 17-of-43   9.8 GB: 18-of-43   9.9 GB: 19-of-43   9.9 GB: 20-of-43   9.8 GB: 21-of-43   9.7 GB: 22-of-43   9.8 GB: 23-of-43   9.9 GB: 24-of-43   9.8 GB: 25-of-43   10.0 GB: 26-of-43   9.8 GB: 27-of-43   9.9 GB: 28-of-43   9.7 GB: 29-of-43   9.9 GB: 30-of-43   9.8 GB: 31-of-43   10.0 GB: 32-of-43   10.0 GB: 33-of-43   10.0 GB: 34-of-43   10.0 GB: 35-of-43   9.8 GB: 36-of-43   9.8 GB: 37-of-43   9.9 GB: 38-of-43   9.7 GB: 39-of-43   9.8 GB: 40-of-43   9.8 GB: 41-of-43   10.0 GB: 42-of-43   3.7 GB: 43-of-43
Supported Languagesen
Model ArchitectureLlamaForCausalLM
Licensellama2
Context Length4096
Model Max Length4096
Transformers Version4.36.1
Vocabulary Size32000
Torch Data Typefloat16

Quantized Models of the Saily 220B

Model
Likes
Downloads
VRAM
Saily 220B GGUF970 GB
Saily 220B GPTQ14105 GB
Saily 220B AWQ05109 GB

Best Alternatives to Saily 220B

Best Alternatives
Context / RAM
Downloads
Likes
...l Llama 220M GQA 32K Theta Sft32K / 0.4 GB62
Smol Llama 220M GQA 32K Theta32K / 0.4 GB81
Nuevamol 220M Base8K / 0.4 GB60
Smol Llama 220M GQA2K / 0.4 GB32613
Smol Llama 220M Openhermes2K / 0.4 GB1135
Smol Llama 220M Open Instruct2K / 0.4 GB62
...mol Llama 220M GQA Fineweb Edu2K / 0.4 GB141
Smol Llama 220M Bees Internal2K / 0.4 GB61
Beecoder 220M Python2K / 0.4 GB123
Saily 220B GPTQ4K / 105.2 GB41
Note: green Score (e.g. "73.2") means that the model is better than deepnight-research/Saily_220B.