Lernex
/

Metis-1.5-base

Text Generation

metis_multihead_latent_moe

mixture-of-experts

Mixture of Experts

Model card Files Files and versions

Metis-1.5-base / export_meta.json

GiuliannoV's picture

Metis-1.5-base: 898M single-latent MoE, 50B-token JAX/TPU pretrain

497f6ee verified 23 days ago

History Blame Contribute Delete

132 Bytes

	{
	"params": 898051168,
	"source_step": 95367,
	"dtype": "bfloat16",
	"n_tensors": 268,
	"format": "metis_jax_safetensors_v1"
	}