mradermacher/MasriSwitch-Gemma3n-Transcriber-v1-GGUF

Same weights, converted (fewer layers) vs oddadmix/MasriSwitch-Gemma3n-Transcriber-v1. mradermacher/MasriSwitch-Gemma3n-Transcriber-v1-GGUF matches part of oddadmix/MasriSwitch-Gemma3n-Transcriber-v1 (809 of its 1676 tensors; the rest of oddadmix/MasriSwitch-Gemma3n-Transcriber-v1 has no counterpart). On the shared part: Same weights stored in another dtype (BF16 → F16/BF16), plus 37 new tensors.

Tensors changed: Embeddings 0/1, other 0/246, Norms 0/317, MLP 0/105, Attention 0/140

On the Hub · Downloads on Model Pulse