Metrics

PPL	arc_easy	arc_challenge	piqa	winogrande	hellaswag	mmlu	QA Avg
4082.93	26.26 ± 0.90	22.44 ± 1.22	52.45 ± 1.17	51.78 ± 1.40	25.88 ± 0.44	-	35.76

Training method based on BitDistiller Paper

Safetensors

Model size

7B params

Tensor type

BF16

Inference Providers NEW

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

Model tree for Heisenger/Llama-2-7b-hf_1bit_int

Base model

Finetuned

(53)

this model