LLM EXPLORER 59,358 MODELS INDEXED

Tinyllama Zh by whynlp

By whynlp · 8 downloads
HF Repo ↗ Copy URL

Tinyllama Zh is an open-source language model by whynlp. Features: 90b LLM, VRAM: 4.9GB, Context: 2K, License: mit, LLM Explorer Score: 0.11.

  Conversational   Dataset:p208p2002/wudao   Endpoints compatible   Llama   Region:us   Safetensors   Zh
Model Card on HF 🤗: https://huggingface.co/whynlp/tinyllama-zh 

Tinyllama Zh Parameters and Internals

Model Type 
Llama, Causal Language Model
Use Cases 
Areas:
demo project, pretraining
Primary Use Cases:
demonstrating how to use the tinyllama to pretrain on a large corpus
Limitations:
does not perform very well
Considerations:
For better performance, use a better corpus such as wanjuan.
Additional Notes 
This project serves as a demonstration to pretrain the TinyLlama on a large corpus with minimal modification to the transformers code.
Supported Languages 
Chinese (NLP)
Training Details 
Data Sources:
WuDaoCorpora Text
Data Volume:
45B tokens
Training Time:
6 days
Hardware Used:
8 A100 GPUs
LLM NameTinyllama Zh
Repository 🤗https://huggingface.co/whynlp/tinyllama-zh 
Model Size90b
Required VRAM4.9 GB
Updated2026-07-10
Maintainerwhynlp
Model Typellama
Model Files  4.9 GB
Supported Languageszh
Model ArchitectureLlamaForCausalLM
Licensemit
Context Length2048
Model Max Length2048
Transformers Version4.35.2
Tokenizer ClassChatGLMTokenizer
Padding Token<unk>
Vocabulary Size65024
Torch Data Typefloat32

Best Alternatives to Tinyllama Zh

Best Alternatives
Context / RAM
Downloads
Likes
Chytrej 90M Base8K / 0.2 GB3022
Chytrej1.5 90M Base8K / 0.2 GB112
Chytrej1 90M Base8K / 0.2 GB162
BigWeave V6 90B4K / 175.7 GB830
BigWeave V12 90B4K / 175.7 GB892
BigWeave V9 90B4K / 175.7 GB41
Note: green Score (e.g. "73.2") means that the model is better than whynlp/tinyllama-zh.