What are the hardware requirements for Lamma3merge2 15B MoE?

Lamma3merge2 15B MoE requires approximately 27.5 GB of VRAM and supports a context window of 8K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Lamma3merge2 15B MoE and how large is it?

Lamma3merge2 15B MoE is developed by allknowingroger, a model with 8b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Lamma3merge2 15B MoE perform on standard benchmarks?

Lamma3merge2 15B MoE has the following published scores: MMLU 67. Compare against reference models on this page or on the LLM Explorer leaderboards.

Lamma3merge2 15B MoE by allknowingroger — VRAM 27.5GB, 8K context

Name: Lamma3merge2 15B MoE
Author: allknowingroger

Lamma3merge2 15B MoE is an open-source language model by allknowingroger. Features: 8b LLM, VRAM: 27.5GB, Context: 8K, License: apache-2.0, Instruction-Based, LLM Explorer Score: 0.13, Arc: 60, HellaSwag: 79.5, MMLU: 67, GSM8K: 67.9.

Autotrain compatible Base model:birgermoell/llama-3... Base model:finetune:birgermoel... Birgermoell/llama-3-dare ties Conversational Endpoints compatible Frankenmoe Instruct Lazymergekit Merge Mergekit Mixtral Moe Mohsenfayyaz/meta-llama-3-8b-i... Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/allknowingroger/Lamma3merge2-15B-MoE

Lamma3merge2 15B MoE Parameters and Internals

Model Type

Mixture of Experts (MoE)

Additional Notes

Lamma3merge2-15B-MoE is a Mixture of Experts model created using LazyMergekit, combining two separate models.

LLM Name	Lamma3merge2 15B MoE
Repository 🤗	https://huggingface.co/allknowingroger/Lamma3merge2-15B-MoE
Base Model(s)	mohsenfayyaz/Meta-Llama-3-8B-Instruct_esnli_5000_1ep Llama 3 Dare Ties mohsenfayyaz/Meta-Llama-3-8B-Instruct_esnli_5000_1ep birgermoell/Llama-3-dare_ties
Model Size	8b
Required VRAM	27.5 GB
Updated	2024-12-03
Maintainer	allknowingroger
Model Type	mixtral
Instruction-Based	Yes
Model Files	1.1 GB: 1-of-15 2.0 GB: 2-of-15 2.0 GB: 3-of-15 2.0 GB: 4-of-15 2.0 GB: 5-of-15 2.0 GB: 6-of-15 2.0 GB: 7-of-15 2.0 GB: 8-of-15 2.0 GB: 9-of-15 2.0 GB: 10-of-15 2.0 GB: 11-of-15 2.0 GB: 12-of-15 2.0 GB: 13-of-15 2.0 GB: 14-of-15 0.4 GB: 15-of-15
Model Architecture	MixtralForCausalLM
License	apache-2.0
Context Length	8192
Model Max Length	8192
Transformers Version	4.40.0
Tokenizer Class	PreTrainedTokenizerFast
Padding Token	<\|begin_of_text\|>
Vocabulary Size	128256
Torch Data Type	bfloat16

Rank the Lamma3merge2 15B MoE 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 Lamma3merge2 15B MoE

Best Alternatives	Context / RAM	Downloads	Likes
Lamma3merge3 15B MoE	8K / 27.5 GB	11	1
Mergkit 1	8K / 22.6 GB	5	0
Llama3merge8 15B MoE	8K / 27.5 GB	5	0

Note: green Score (e.g. "73.2") means that the model is better than allknowingroger/Lamma3merge2-15B-MoE.