What are the hardware requirements for Saiga2 13B Lora?

Saiga2 13B Lora requires approximately 0.1 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Saiga2 13B Lora and how large is it?

Saiga2 13B Lora is developed by IlyaGusev, a model with 13b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Saiga2 13B Lora?

Saiga2 13B Lora is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Saiga2 13B Lora by IlyaGusev — VRAM 0.1GB

Name: Saiga2 13B Lora
Rating: 4 (3 reviews)
Author: IlyaGusev

Saiga2 13B Lora is an open-source language model by IlyaGusev. Features: 13b LLM, VRAM: 0.1GB, License: cc-by-4.0, Instruction-Based, LLM Explorer Score: 0.08.

Adapter Conversational Dataset:ilyagusev/oasst1 ru ma... Dataset:ilyagusev/ru sharegpt ... Dataset:ilyagusev/ru turbo alp... Dataset:ilyagusev/ru turbo alp... Dataset:ilyagusev/ru turbo sai... Dataset:lksy/ru instruct gpt4 Finetuned Instruct Lora Region:us Ru

Model Card on HF 🤗: https://huggingface.co/IlyaGusev/saiga2_13b_lora

Saiga2 13B Lora Parameters and Internals

Model Type

conversational

Additional Notes

Based on LLaMA-2 13B HF with an adapter-only version for Russian language.

Supported Languages

ru (Russian)

Training Details

Data Sources:

IlyaGusev/ru_turbo_alpaca, IlyaGusev/ru_turbo_saiga, IlyaGusev/ru_sharegpt_cleaned, IlyaGusev/oasst1_ru_main_branch, IlyaGusev/ru_turbo_alpaca_evol_instruct, lksy/ru_instruct_gpt4

Methodology:

Adapter-only version trained using the development version of `transformers` and `peft`.

Model Architecture:

LLaMA-2 based chatbot

Input Output

Input Format:

~~{role} {content}~~

Accepted Modalities:

text

Output Format:

text

Release Notes

Version:

Notes:

Trained on multiple datasets including ru_turbo_alpaca and ru_turbo_saiga.

LLM Name	Saiga2 13b Lora
Repository 🤗	https://huggingface.co/IlyaGusev/saiga2_13b_lora
Model Size	13b
Required VRAM	0.1 GB
Updated	2026-07-26
Maintainer	IlyaGusev
Instruction-Based	Yes
Model Files	0.1 GB
Supported Languages	ru
Model Architecture	Adapter
License	cc-by-4.0
Model Max Length	2048
Is Biased	none
Tokenizer Class	LlamaTokenizer
PEFT Type	LORA
LoRA Model	Yes
PEFT Target Modules	q_proj\|v_proj\|k_proj\|o_proj
LoRA Alpha	16
LoRA Dropout	0.05
R Param	16

Rank the Saiga2 13B Lora capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

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

Best Alternatives to Saiga2 13B Lora

Best Alternatives	Context / RAM	Downloads	Likes
...10 2024 06 23 06 24 24 3558635	0K / 0.8 GB	5	0
...10 2024 06 22 21 11 23 3558624	0K / 0.8 GB	5	0
Typescript Instruct 20K V4	0K / 26 GB	67	2
Llama 2 13B Instruct V0.2	0K / 0.2 GB	11	10
Typescript Instruct 20K V2	0K / 26 GB	48	2
... 13B Instruct Lora Jaster V1.0	0K / 0.1 GB	13	2
...t Lora Jaster Dolly Oasst V1.0	0K / 0.1 GB	6	1
...Instruct Lora Dolly Oasst V1.0	0K / 0.1 GB	2	1
RuGPT 3.5 13B Lora	0K / 0.1 GB	7	12
...lama 13B Instruction Finetune2	0K / 0 GB	10	1

Note: green Score (e.g. "73.2") means that the model is better than IlyaGusev/saiga2_13b_lora.