Metrics

PPL	arc_easy	arc_challenge	piqa	winogrande	hellaswag	mmlu	QA Avg
16.94	36.91 ± 1.00	20.14 ± 1.00	60.28 ± 1.00	53.99 ± 1.00	33.00 ± 1.00	-	40.86

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 VictorFiz/TinyLlama_v1.1_2bit_int_llama2_7b_teacher

Base model

Finetuned

(53)

this model