gptkbp:instanceOf
|
large language model
|
gptkbp:architecture
|
gptkb:transformation
|
gptkbp:bench
|
outperforms Llama-2 13B on many tasks
|
gptkbp:compatibleWith
|
gptkb:microprocessor
gptkb:NVIDIA_GPUs
|
gptkbp:context
|
8,192 tokens
|
gptkbp:developer
|
gptkb:Mistral_AI
|
gptkbp:github
|
https://github.com/mistralai/mistral-src
|
gptkbp:hasModel
|
decoder-only
|
https://www.w3.org/2000/01/rdf-schema#label
|
Mistral-7B
|
gptkbp:intendedUse
|
chatbot
translator
code generation
summarization
text generation
reasoning tasks
|
gptkbp:language
|
gptkb:French
English
|
gptkbp:license
|
Apache 2.0
|
gptkbp:memoryRequirement
|
~13GB (FP16)
~7GB (INT4)
|
gptkbp:modelCard
|
https://huggingface.co/mistralai/Mistral-7B-v0.1
|
gptkbp:notableFeature
|
grouped-query attention
efficient inference
sliding window attention
|
gptkbp:notableFor
|
commercial applications
open-source research
|
gptkbp:openSource
|
true
|
gptkbp:parameter
|
7 billion
|
gptkbp:pretrainedWeightsAvailable
|
true
|
gptkbp:quantizedVersionsAvailable
|
true
|
gptkbp:releaseDate
|
2023-09-27
|
gptkbp:successor
|
gptkb:Mixtral-8x7B
Mistral-8x7B
|
gptkbp:tokenizer
|
custom tokenizer
|
gptkbp:trainer
|
gptkb:law
books
web data
|
gptkbp:trainingObjective
|
causal language modeling
|
gptkbp:bfsParent
|
gptkb:Hugging_Face_Hub
gptkb:Mistral_API
|
gptkbp:bfsLayer
|
7
|