anemll's picture
Version 0.1.1 for Prefill - (1,B,H) - incorrect!
62682fa verified
raw
history blame
565 Bytes
model_info:
name: anemll-Meta-Llama-3.2-1B-LUT8-ctx512
version: 0.1.1
description: |
Demonstarates running Meta-Llama-3.2-1B on Apple Neural Engine
Context length: 512
Batch size: 64
Chunks: 2
license: MIT
author: Anemll
framework: Core ML
language: Python
parameters:
context_length: 512
batch_size: 64
lut_embeddings: 8
lut_ffn: 8
lut_lmhead: 8
num_chunks: 2
model_prefix: llama
embeddings: llama_embeddings_lut8.mlmodelc
lm_head: llama_lm_head_lut8.mlmodelc
ffn: llama_FFN_PF_lut8.mlmodelc