meta.yaml

model_info:
  name: anemll-meta-llama-Llama-3.2-1B-Instruct-ctx1024
  version: 0.3.5
  description: |
    Demonstarates running meta-llama-Llama-3.2-1B-Instruct on Apple Neural Engine
    Context length: 1024
    Batch size: 64
    Chunks: 1
  license: MIT
  author: Anemll
  framework: Core ML
  language: Python
  architecture: llama
  parameters:
    context_length: 1024
    batch_size: 64
    lut_embeddings: none
    lut_ffn: 6
    lut_ffn_per_channel: 4
    lut_lmhead: 6
    lut_lmhead_per_channel: 4
    num_chunks: 1
    model_prefix: llama
    embeddings: llama_embeddings.mlmodelc
    lm_head: llama_lm_head_lut6.mlmodelc
    ffn: llama_FFN_PF_lut6_chunk_01of01.mlmodelc
    split_lm_head: 8
    argmax_in_model: true
    vocab_size: 128256
    lm_head_chunk_sizes: [16032, 16032, 16032, 16032, 16032, 16032, 16032, 16032]
    prefill_dynamic_slice: true

# =============================================================================
# Conversion Parameters (for troubleshooting)
# =============================================================================
# Generated: 2026-02-12 10:08:46
#
# model_path: /Users/anemll/.cache/huggingface/hub/models--meta-llama--Llama-3.2-1B-Instruct/snapshots/9213176726f574b556790deb65791e0c5aa438b6
# output_dir: /Users/anemll/Models/ANE/llama3.2-1b-instruct-ctx1024
# command_line: "./anemll/utils/convert_model.sh --model meta-llama/Llama-3.2-1B-Instruct --output /Users/anemll/Models/ANE/llama3.2-1b-instruct-ctx1024 --context 1024 --lut2 6\\,4 --lut3 6\\,4 --chunk 1 --argmax"
# context_length: 1024
# batch_size: 64
# num_chunks: 1
# lut_part1: none
# lut_part2: 6,4
# lut_part3: 6,4
# prefix: llama
# architecture: llama
# argmax_in_model: true
# split_rotate: false
# single_cache: false
# dynamic_prefill_slice: true
# monolithic: false
# anemll_version: 0.3.5
# vocab_size: 128256
# lm_head_chunk_sizes: "[16032, 16032, 16032, 16032, 16032, 16032, 16032, 16032]"
# =============================================================================