model used: microsoft/Phi-3-mini-128k-instruct
LASER scripts used: rmt_laser.py,rmt_laser_snr.py,rmt_laser_snr_math.py from laserRMT

perplexity is reduced comparing to base model, waiting for HF eval results - due to trust_remote_code=True it won't happen soon.

Safetensors

Model size

3.82B params

Tensor type

BF16

Inference Providers NEW

This model is not currently available via any of the supported Inference Providers.

Collection including TeeZee/Phi-3-mini-128k-instruct-LASER