model_info:
  name: anemll-WeiboAI-VibeThinker-1.5B-ctx2048
  version: 0.3.5
  description: 'Demonstarates running WeiboAI-VibeThinker-1.5B on Apple Neural Engine

    Context length: 2048

    Batch size: 64

    Chunks: 4

    '
  license: MIT
  author: Anemll
  framework: Core ML
  language: Python
  architecture: qwen2
  parameters:
    context_length: 2048
    batch_size: 64
    lut_embeddings: none
    lut_ffn: 6
    lut_ffn_per_channel: 4
    lut_lmhead: 6
    lut_lmhead_per_channel: 4
    num_chunks: 4
    model_prefix: qwen25
    embeddings: qwen25_embeddings.mlmodelc
    lm_head: qwen25_lm_head_lut6.mlmodelc
    ffn: qwen25_FFN_PF_lut6_chunk_01of04.mlmodelc
    split_lm_head: 16
    vocab_size: 151936
    lm_head_chunk_sizes:
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    - 9496
    prefill_dynamic_slice: true
    argmax_in_model: false
    recommended_sampling:
      do_sample: true
      temperature: 0.6
      top_p: 0.95
      top_k: 0