int8/config.json

{
  "hidden_size": 1024,
  "num_hidden_layers": 24,
  "num_attention_heads": 8,
  "num_key_value_heads": 2,
  "head_dim": 256,
  "intermediate_size": 3584,
  "vocab_size": 248320,
  "max_seq_len": 2048,
  "rope_theta": 10000000,
  "rms_norm_eps": 1e-06,
  "eos_token_id": 248046,
  "pad_token_id": 248044,
  "quantization": "int8",
  "quantization_bits": 8,
  "quantization_group_size": 64,
  "model_type": "qwen3_5_text",
  "layer_types": [
    "linear_attention",
    "linear_attention",
    "linear_attention",
    "full_attention",
    "linear_attention",
    "linear_attention",
    "linear_attention",
    "full_attention",
    "linear_attention",
    "linear_attention",
    "linear_attention",
    "full_attention",
    "linear_attention",
    "linear_attention",
    "linear_attention",
    "full_attention",
    "linear_attention",
    "linear_attention",
    "linear_attention",
    "full_attention",
    "linear_attention",
    "linear_attention",
    "linear_attention",
    "full_attention"
  ],
  "full_attention_interval": 4,
  "linear_num_key_heads": 16,
  "linear_key_head_dim": 128,
  "linear_num_value_heads": 16,
  "linear_value_head_dim": 128,
  "linear_conv_kernel_dim": 4,
  "partial_rotary_factor": 0.25,
  "tie_word_embeddings": true
}