Metrics

PPL	arc_easy	arc_challenge	piqa	winogrande	hellaswag	mmlu	QA Avg
11.52	43.60 ± 1.02	23.46 ± 1.24	66.38 ± 1.10	52.01 ± 1.40	38.80 ± 0.49	-	44.85

Training method based on BitDistiller Paper

Safetensors

Model size

1B params

Tensor type

BF16

Inference Providers NEW

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

Model tree for fredericowieser/TinyLlama_v1.1_3bit_nf3

Base model

Finetuned

(53)

this model