LLM EXPLORER 56,604 MODELS INDEXED

EEVE Korean 2.8B V1.0 by yanolja

By yanolja · 1659 downloads

EEVE Korean 2.8B V1.0 is an open-source language model by yanolja. Features: 2.8b LLM, VRAM: 5.7GB, Context: 2K, License: apache-2.0, LLM Explorer Score: 0.15, Arc: 57.3, HellaSwag: 72.2, MMLU: 51.6, GSM8K: 36.4.

  Arxiv:2402.14714   Autotrain compatible Base model:finetune:microsoft/...   Base model:microsoft/phi-2   Conversational   Custom code   Endpoints compatible   Generated from trainer   Phi   Region:us   Safetensors   Sharded   Tensorflow

EEVE Korean 2.8B V1.0 Parameters and Internals

Model Type 
Multilingual, Language Model
Use Cases 
Primary Use Cases:
Korean language tasks
Limitations:
Model hasn't been fine-tuned with instruction-based training
Additional Notes 
The model is fine-tuned specifically for Korean tasks using an extended Korean vocabulary. It expands upon the English-trained base model by adding new Korean tokens and partially fine-tuning relevant embeddings.
Supported Languages 
Korean (fluent)
Training Details 
Data Sources:
Korean web-crawled datasets
Methodology:
Subword-based embedding with parameter freezing
LLM NameEEVE Korean 2.8B V1.0
Repository πŸ€—https://huggingface.co/yanolja/EEVE-Korean-2.8B-v1.0 
Base Model(s)  Phi 2   microsoft/phi-2
Model Size2.8b
Required VRAM5.7 GB
Updated2025-09-02
Maintaineryanolja
Model Typephi
Model Files  5.0 GB: 1-of-2   0.7 GB: 2-of-2
Model ArchitecturePhiForCausalLM
Licenseapache-2.0
Context Length2048
Model Max Length2048
Transformers Version4.38.0.dev0
Tokenizer ClassLlamaTokenizer
Padding Token</s>
Vocabulary Size58944
Torch Data Typebfloat16

Best Alternatives to EEVE Korean 2.8B V1.0

Best Alternatives
Context / RAM
Downloads
Likes
MFANN3bv0.24128K / 11.1 GB60
MFANN3b128K / 11.1 GB860
MFANN3bv1.4128K / 11.1 GB80
MFANN Phigments Slerp V3.2128K / 5.6 GB60
MFANN3bv1.3128K / 11.1 GB140
MFANN3bv0.23128K / 11.1 GB100
MFANN3bv1.1128K / 11.1 GB100
MFANN3bv1.5128K / 11.1 GB80
MFANN Liminerity Slerp 4a128K / 5.6 GB70
MFANN3b Rebase128K / 11.1 GB70
Note: green Score (e.g. "73.2") means that the model is better than yanolja/EEVE-Korean-2.8B-v1.0.