Meditron 7B By epfl-llm: Benchmarks, Features and Detailed Analysis. Insights on Meditron 7B.

Arxiv:2311.16079 Autotrain compatible Base model:finetune:meta-llama... Base model:meta-llama/llama-2-... Dataset:epfl-llm/guidelines En Endpoints compatible Llama Pytorch Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/epfl-llm/meditron-7b

Meditron 7B Benchmarks

LLME Score: 0.18627

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

What is the LLM Explorer Rank (Score)

Meditron 7B Parameters and Internals

Model Type

Causal decoder-only transformer language model

Use Cases

Areas:

Medical exam question answering, Supporting differential diagnosis, Disease information query, General health information query

Limitations:

Not recommended for production use, Unsuitable for professional purposes related to health and medicine

Considerations:

Use in production environments requires rigorous evaluation and alignment processes.

Additional Notes

Significant research is still required to fully explore potential bias, fairness, and safety issues.

Supported Languages

English (mainly)

Training Details

Data Sources:

Clinical Guidelines, Medical Paper Abstracts, Medical Papers, Replay Data

Data Volume:

48.1B tokens

Methodology:

Continued pretraining

Context Length:

2048

Training Time:

September 2023

Hardware Used:

1 node of 8x NVIDIA A100 (80GB) SXM GPUs

Model Architecture:

Llama 2, Hidden dimension: 4096, Number of attention heads: 32, Number of layers: 32

Input Output

Input Format:

Text-only data

Accepted Modalities:

Text

Output Format:

Text

LLM Name	Meditron 7B
Repository 🤗	https://huggingface.co/epfl-llm/meditron-7b
Base Model(s)	Llama 2 7B meta-llama/Llama-2-7b
Model Size	7b
Required VRAM	13.4 GB
Updated	2025-02-22
Maintainer	epfl-llm
Model Type	llama
Model Files	1.9 GB: 1-of-8 1.9 GB: 2-of-8 1.8 GB: 3-of-8 1.9 GB: 4-of-8 1.9 GB: 5-of-8 1.8 GB: 6-of-8 1.9 GB: 7-of-8 0.3 GB: 8-of-8 1.9 GB: 1-of-8 1.9 GB: 2-of-8 1.8 GB: 3-of-8 1.9 GB: 4-of-8 1.9 GB: 5-of-8 1.8 GB: 6-of-8 1.9 GB: 7-of-8 0.3 GB: 8-of-8
Supported Languages	en
Gated Model	Yes
Model Architecture	LlamaForCausalLM
License	proprietary
Context Length	2048
Model Max Length	2048
Transformers Version	4.35.2
Tokenizer Class	LlamaTokenizer
Padding Token	<PAD>
Vocabulary Size	32017
Torch Data Type	bfloat16

Quantized Models of the Meditron 7B

Model	Likes	Downloads	VRAM
...editron 7B Lora Finetuned 4bit	0	22	3 GB
Meditron 7B AWQ	2	9046	3 GB
Meditron 7B GGUF	21	694	2 GB
Meditron 7B GPTQ	3	40	3 GB

Best Alternatives to Meditron 7B

Best Alternatives	Context / RAM	Downloads	Likes
2 Very Sci Fi	1024K / 16.1 GB	317	0
...1M 1000000ctx AEZAKMI 3 1 1702	1024K / 13.5 GB	23	1
... Qwen2.5llamaify 7B V23.1 200K	195K / 15.2 GB	3943	3
LlamaStock 8B	128K / 16.1 GB	11	1
SuperNeuralDreadDevil 8B	128K / 16.1 GB	54	1
Yarn Llama 2 7B 128K	128K / 13.5 GB	6422	39
LLaMA 7B PoSE YaRN 128K	128K / 13.5 GB	7	3
LLaMA 7B PoSE Linear 96K	96K / 27 GB	9	2
LLaMA 7B PoSE YaRN 96K	96K / 13.5 GB	11	1
Chat Llama2 7B 80K	80K / 13.8 GB	8	0

Note: green Score (e.g. "73.2") means that the model is better than epfl-llm/meditron-7b.

Rank the Meditron 7B Capabilities

🆘 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! 🌟

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? 43470 in total.

Email us: info@extractum.io. Our Privacy Policy | Terms and Conditions | Suggest an improvement.

Our Social Media →

Original data from HuggingFace, OpenCompass and various public git repos.

Release v20241227

Support LLM Explorer