APT3 500M Base by Azurro

 ยป  All LLMs  ยป  Azurro  ยป  APT3 500M Base   URL Share it on

  Allamo   Autotrain compatible Dataset:chrisociepa/wikipedia-...   Llama   Pl   Region:us   Safetensors
Model Card on HF ๐Ÿค—: https://huggingface.co/Azurro/APT3-500M-Base 

APT3 500M Base Benchmarks

nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
APT3 500M Base (Azurro/APT3-500M-Base)

APT3 500M Base Parameters and Internals

Model Type 
causal decoder-only
Use Cases 
Areas:
Research, Non-commercial applications
Limitations:
Not intended for deployment without fine-tuning. Not for human-facing interactions without further guardrails and user consent. Can produce factually incorrect outputs.
Additional Notes 
APT3-500M-Base is trained with Polish corpus prioritizing high-quality, manually selected texts over large-scale, lower quality datasets.
Supported Languages 
pl (native)
Training Details 
Data Sources:
ebooks, Polish Wikipedia, web crawl data
Data Volume:
Over 20 billion tokens
Context Length:
1024
Hardware Used:
Nvidia's RTX 4090 24GB VRAM
Model Architecture:
Transformer-based, similar to Meta AIโ€™s LLaMA
Input Output 
Performance Tips:
Use smaller precision (`bfloat16`) to reduce memory usage.
LLM NameAPT3 500M Base
Repository ๐Ÿค—https://huggingface.co/Azurro/APT3-500M-Base 
Model Size500m
Required VRAM1.9 GB
Updated2025-02-14
MaintainerAzurro
Model Typellama
Model Files  0.0 GB   1.9 GB   3.8 GB   1.9 GB
Supported Languagespl
Model ArchitectureLlamaForCausalLM
Licensecc-by-nc-4.0
Context Length2048
Model Max Length2048
Transformers Version4.35.0
Tokenizer ClassPreTrainedTokenizerFast
Vocabulary Size31980
Torch Data Typefloat32

Best Alternatives to APT3 500M Base

Best Alternatives
Context / RAM
Downloads
Likes
H2o Danube3 500M Chat8K / 1 GB436932
Acheron D8K / 1 GB80
Text Arco8K / 1 GB1730
Acheron C8K / 1 GB90
Acheron8K / 1 GB190
Arco Plus8K / 1.6 GB1386
H2o Danube3 500M Base8K / 1 GB85229
Danube Reasoner8K / 1 GB1340
Massive8K / 1 GB1140
Clite 500M8K / 1 GB203
Note: green Score (e.g. "73.2") means that the model is better than Azurro/APT3-500M-Base.

Rank the APT3 500M Base Capabilities

๐Ÿ†˜ Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! ๐ŸŒŸ

Instruction Following and Task Automation  
Factuality and Completeness of Knowledge  
Censorship and Alignment  
Data Analysis and Insight Generation  
Text Generation  
Text Summarization and Feature Extraction  
Code Generation  
Multi-Language Support and Translation  

What open-source LLMs or SLMs are you in search of? 43106 in total.

Our Social Media →  
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241227