needle-mlx / config.json
TiGa-RCE's picture
Preserve Needle bfloat16 runtime behavior
64bce88 verified
Raw
History Blame Contribute Delete
717 Bytes
{
"vocab_size": 8192,
"d_model": 512,
"num_heads": 8,
"num_kv_heads": 4,
"num_encoder_layers": 12,
"num_decoder_layers": 8,
"d_ff": 2048,
"max_seq_len": 1024,
"pad_token_id": 0,
"rope_theta": 10000.0,
"dtype": "bfloat16",
"activation": "swiglu",
"num_memory_slots": 64,
"n_mels": 80,
"dropout_rate": 0.1,
"contrastive_dim": 128,
"enable_speech": false,
"no_feedforward": true,
"model_type": "needle_mlx",
"architectures": [
"NeedleModel"
],
"format": "mlx-safetensors",
"format_version": 2,
"tensor_dtype": "bfloat16",
"eos_token_id": 1,
"source_checkpoint": "needle.pkl",
"source_sha256": "40a32e91d1d4197bf15ba559b74f6727c342dc8746918742fc7d8e2c1f18df40"
}