Metrics

PPL	arc_easy	arc_challenge	piqa	winogrande	hellaswag	mmlu	QA Avg
8817.78	24.66 ± 0.88	21.84 ± 1.21	53.16 ± 1.16	49.64 ± 1.41	25.96 ± 0.44	-	35.05

Training method based on BitDistiller Paper

Safetensors

Model size

1.1B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for BrownianNotion/TinyLlama_v1.1_1.58bit_ternary

Base model

Finetuned

(54)

this model