Metrics

PPL	arc_easy	arc_challenge	piqa	winogrande	hellaswag	mmlu	QA Avg
22.39	36.11 ± 0.99	21.84 ± 1.21	60.72 ± 1.14	51.38 ± 1.40	32.73 ± 0.47	-	40.56

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 Heisenger/TinyLlama_v1.1_2bit_int_three_times_data

Base model

Finetuned

(53)

this model