Towards Cross-Tokenizer Distillation: the Universal Logit Distillation Loss for LLMs
Paper
•
2402.12030
•
Published
LLMs Distillation
-
-
mistralai/Mistral-7B-Instruct-v0.2
Text Generation • Updated • 4.76M • • 2.61k -
meta-llama/Llama-2-7b-chat-hf
Text Generation • Updated • 1.1M • • 4.12k -
EleutherAI/pythia-160m-deduped
Text Generation • Updated • 25k • 3 -
EleutherAI/pythia-410m-deduped
Text Generation • Updated • 7.03k • 20 -
EleutherAI/pythia-1b-deduped
Text Generation • Updated • 13.9k • 17 -
bigscience/bloomz-560m
Text Generation • Updated • 4.76M • 114 -
bigscience/mt0-base
Text2Text Generation • Updated • 989 • 30 -
facebook/opt-350m
Text Generation • Updated • 159k • 131 -
rajpurkar/squad
Viewer • Updated • 98.2k • 47.7k • 281 -
google-research-datasets/qed
Updated • 91 • 4 -
GEM/FairytaleQA
Viewer • Updated • 10.6k • 244 • 8 -
qiaojin/PubMedQA
Viewer • Updated • 274k • 6.92k • 165 -
knkarthick/dialogsum
Viewer • Updated • 14.5k • 10k • 191