Metrics

PPL	arc_easy	arc_challenge	piqa	winogrande	hellaswag	mmlu	QA Avg
2203.2	25.38 ± 0.89	22.35 ± 1.22	53.05 ± 1.16	49.80 ± 1.41	25.76 ± 0.44	-	35.27

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_range_scale_4epochs

Base model

Finetuned

(54)

this model