text_encoder/quantization_config.json

{
  "add_skip_keys": true,
  "codebook_steps": 24,
  "dequantize_fp32": true,
  "dynamic_loss_threshold": null,
  "group_size": 64,
  "hadamard_group_size": 256,
  "is_integer": true,
  "is_training": false,
  "is_unsigned": true,
  "minimum_allowed_channel_size": 32,
  "minimum_allowed_numel": 16384,
  "modules_dtype_dict": {},
  "modules_quant_config": {},
  "modules_to_not_convert": [
    "lm_head",
    "embedding_projection",
    "embed_tokens",
    "model.language_model.embed_tokens.weight",
    "model.language_model.layers.0.input_layernorm.weight",
    "model.language_model.layers.0.post_attention_layernorm.weight",
    "model.language_model.layers.0.self_attn.k_norm.weight",
    "model.language_model.layers.0.self_attn.q_norm.weight",
    "model.language_model.layers.1.input_layernorm.weight",
    "model.language_model.layers.1.post_attention_layernorm.weight",
    "model.language_model.layers.1.self_attn.k_norm.weight",
    "model.language_model.layers.1.self_attn.q_norm.weight",
    "model.language_model.layers.2.input_layernorm.weight",
    "model.language_model.layers.2.post_attention_layernorm.weight",
    "model.language_model.layers.2.self_attn.k_norm.weight",
    "model.language_model.layers.2.self_attn.q_norm.weight",
    "model.language_model.layers.3.input_layernorm.weight",
    "model.language_model.layers.3.post_attention_layernorm.weight",
    "model.language_model.layers.3.self_attn.k_norm.weight",
    "model.language_model.layers.3.self_attn.q_norm.weight",
    "model.language_model.layers.4.input_layernorm.weight",
    "model.language_model.layers.4.post_attention_layernorm.weight",
    "model.language_model.layers.4.self_attn.k_norm.weight",
    "model.language_model.layers.4.self_attn.q_norm.weight",
    "model.language_model.layers.5.input_layernorm.weight",
    "model.language_model.layers.5.post_attention_layernorm.weight",
    "model.language_model.layers.5.self_attn.k_norm.weight",
    "model.language_model.layers.5.self_attn.q_norm.weight",
    "model.language_model.layers.6.input_layernorm.weight",
    "model.language_model.layers.6.post_attention_layernorm.weight",
    "model.language_model.layers.6.self_attn.k_norm.weight",
    "model.language_model.layers.6.self_attn.q_norm.weight",
    "model.language_model.layers.7.input_layernorm.weight",
    "model.language_model.layers.7.post_attention_layernorm.weight",
    "model.language_model.layers.7.self_attn.k_norm.weight",
    "model.language_model.layers.7.self_attn.q_norm.weight",
    "model.language_model.layers.8.input_layernorm.weight",
    "model.language_model.layers.8.post_attention_layernorm.weight",
    "model.language_model.layers.8.self_attn.k_norm.weight",
    "model.language_model.layers.8.self_attn.q_norm.weight",
    "model.language_model.layers.10.input_layernorm.weight",
    "model.language_model.layers.10.post_attention_layernorm.weight",
    "model.language_model.layers.10.self_attn.k_norm.weight",
    "model.language_model.layers.10.self_attn.q_norm.weight",
    "model.language_model.layers.11.input_layernorm.weight",
    "model.language_model.layers.11.post_attention_layernorm.weight",
    "model.language_model.layers.11.self_attn.k_norm.weight",
    "model.language_model.layers.11.self_attn.q_norm.weight",
    "model.language_model.layers.12.input_layernorm.weight",
    "model.language_model.layers.12.post_attention_layernorm.weight",
    "model.language_model.layers.12.self_attn.k_norm.weight",
    "model.language_model.layers.12.self_attn.q_norm.weight",
    "model.language_model.layers.13.input_layernorm.weight",
    "model.language_model.layers.13.post_attention_layernorm.weight",
    "model.language_model.layers.13.self_attn.k_norm.weight",
    "model.language_model.layers.13.self_attn.q_norm.weight",
    "model.language_model.layers.9.input_layernorm.weight",
    "model.language_model.layers.9.post_attention_layernorm.weight",
    "model.language_model.layers.9.self_attn.k_norm.weight",
    "model.language_model.layers.9.self_attn.q_norm.weight",
    "model.language_model.layers.14.input_layernorm.weight",
    "model.language_model.layers.14.post_attention_layernorm.weight",
    "model.language_model.layers.14.self_attn.k_norm.weight",
    "model.language_model.layers.14.self_attn.q_norm.weight",
    "model.language_model.layers.15.input_layernorm.weight",
    "model.language_model.layers.15.post_attention_layernorm.weight",
    "model.language_model.layers.15.self_attn.k_norm.weight",
    "model.language_model.layers.15.self_attn.q_norm.weight",
    "model.language_model.layers.16.input_layernorm.weight",
    "model.language_model.layers.16.post_attention_layernorm.weight",
    "model.language_model.layers.16.self_attn.k_norm.weight",
    "model.language_model.layers.16.self_attn.q_norm.weight",
    "model.language_model.layers.17.input_layernorm.weight",
    "model.language_model.layers.17.post_attention_layernorm.weight",
    "model.language_model.layers.17.self_attn.k_norm.weight",
    "model.language_model.layers.17.self_attn.q_norm.weight",
    "model.language_model.layers.18.input_layernorm.weight",
    "model.language_model.layers.18.post_attention_layernorm.weight",
    "model.language_model.layers.18.self_attn.k_norm.weight",
    "model.language_model.layers.18.self_attn.q_norm.weight",
    "model.language_model.layers.19.input_layernorm.weight",
    "model.language_model.layers.19.post_attention_layernorm.weight",
    "model.language_model.layers.19.self_attn.k_norm.weight",
    "model.language_model.layers.19.self_attn.q_norm.weight",
    "model.language_model.layers.20.input_layernorm.weight",
    "model.language_model.layers.20.post_attention_layernorm.weight",
    "model.language_model.layers.20.self_attn.k_norm.weight",
    "model.language_model.layers.20.self_attn.q_norm.weight",
    "model.language_model.layers.21.input_layernorm.weight",
    "model.language_model.layers.21.post_attention_layernorm.weight",
    "model.language_model.layers.21.self_attn.k_norm.weight",
    "model.language_model.layers.21.self_attn.q_norm.weight",
    "model.language_model.layers.22.input_layernorm.weight",
    "model.language_model.layers.22.post_attention_layernorm.weight",
    "model.language_model.layers.22.self_attn.k_norm.weight",
    "model.language_model.layers.22.self_attn.q_norm.weight",
    "model.language_model.layers.23.input_layernorm.weight",
    "model.language_model.layers.23.post_attention_layernorm.weight",
    "model.language_model.layers.23.self_attn.k_norm.weight",
    "model.language_model.layers.23.self_attn.q_norm.weight",
    "model.language_model.layers.24.input_layernorm.weight",
    "model.language_model.layers.24.post_attention_layernorm.weight",
    "model.language_model.layers.24.self_attn.k_norm.weight",
    "model.language_model.layers.24.self_attn.q_norm.weight",
    "model.language_model.layers.25.input_layernorm.weight",
    "model.language_model.layers.25.post_attention_layernorm.weight",
    "model.language_model.layers.25.self_attn.k_norm.weight",
    "model.language_model.layers.25.self_attn.q_norm.weight",
    "model.language_model.layers.26.input_layernorm.weight",
    "model.language_model.layers.26.post_attention_layernorm.weight",
    "model.language_model.layers.26.self_attn.k_norm.weight",
    "model.language_model.layers.26.self_attn.q_norm.weight",
    "model.language_model.layers.27.input_layernorm.weight",
    "model.language_model.layers.27.post_attention_layernorm.weight",
    "model.language_model.layers.27.self_attn.k_norm.weight",
    "model.language_model.layers.27.self_attn.q_norm.weight",
    "model.language_model.layers.28.input_layernorm.weight",
    "model.language_model.layers.28.post_attention_layernorm.weight",
    "model.language_model.layers.28.self_attn.k_norm.weight",
    "model.language_model.layers.28.self_attn.q_norm.weight",
    "model.language_model.layers.29.input_layernorm.weight",
    "model.language_model.layers.29.post_attention_layernorm.weight",
    "model.language_model.layers.29.self_attn.k_norm.weight",
    "model.language_model.layers.29.self_attn.q_norm.weight",
    "model.language_model.layers.30.input_layernorm.weight",
    "model.language_model.layers.30.post_attention_layernorm.weight",
    "model.language_model.layers.30.self_attn.k_norm.weight",
    "model.language_model.layers.30.self_attn.q_norm.weight",
    "model.language_model.layers.31.input_layernorm.weight",
    "model.language_model.layers.31.post_attention_layernorm.weight",
    "model.language_model.layers.31.self_attn.k_norm.weight",
    "model.language_model.layers.31.self_attn.q_norm.weight",
    "model.language_model.layers.32.input_layernorm.weight",
    "model.language_model.layers.32.post_attention_layernorm.weight",
    "model.language_model.layers.32.self_attn.k_norm.weight",
    "model.language_model.layers.32.self_attn.q_norm.weight",
    "model.language_model.layers.33.input_layernorm.weight",
    "model.language_model.layers.33.post_attention_layernorm.weight",
    "model.language_model.layers.33.self_attn.k_norm.weight",
    "model.language_model.layers.33.self_attn.q_norm.weight",
    "model.language_model.layers.34.input_layernorm.weight",
    "model.language_model.layers.34.post_attention_layernorm.weight",
    "model.language_model.layers.34.self_attn.k_norm.weight",
    "model.language_model.layers.34.self_attn.q_norm.weight",
    "model.language_model.layers.35.input_layernorm.weight",
    "model.language_model.layers.35.post_attention_layernorm.weight",
    "model.language_model.layers.35.self_attn.k_norm.weight",
    "model.language_model.layers.35.self_attn.q_norm.weight",
    "model.language_model.layers.36.input_layernorm.weight",
    "model.language_model.layers.36.post_attention_layernorm.weight",
    "model.language_model.layers.36.self_attn.k_norm.weight",
    "model.language_model.layers.36.self_attn.q_norm.weight",
    "model.language_model.layers.37.input_layernorm.weight",
    "model.language_model.layers.37.post_attention_layernorm.weight",
    "model.language_model.layers.37.self_attn.k_norm.weight",
    "model.language_model.layers.37.self_attn.q_norm.weight",
    "model.language_model.layers.38.input_layernorm.weight",
    "model.language_model.layers.38.post_attention_layernorm.weight",
    "model.language_model.layers.38.self_attn.k_norm.weight",
    "model.language_model.layers.38.self_attn.q_norm.weight",
    "model.language_model.layers.39.input_layernorm.weight",
    "model.language_model.layers.39.post_attention_layernorm.weight",
    "model.language_model.layers.39.self_attn.k_norm.weight",
    "model.language_model.layers.39.self_attn.q_norm.weight",
    "model.language_model.layers.40.input_layernorm.weight",
    "model.language_model.layers.40.post_attention_layernorm.weight",
    "model.language_model.layers.40.self_attn.k_norm.weight",
    "model.language_model.layers.40.self_attn.q_norm.weight",
    "model.language_model.layers.41.input_layernorm.weight",
    "model.language_model.layers.41.post_attention_layernorm.weight",
    "model.language_model.layers.41.self_attn.k_norm.weight",
    "model.language_model.layers.41.self_attn.q_norm.weight",
    "model.language_model.layers.42.input_layernorm.weight",
    "model.language_model.layers.42.post_attention_layernorm.weight",
    "model.language_model.layers.42.self_attn.k_norm.weight",
    "model.language_model.layers.42.self_attn.q_norm.weight",
    "model.language_model.layers.43.input_layernorm.weight",
    "model.language_model.layers.43.post_attention_layernorm.weight",
    "model.language_model.layers.43.self_attn.k_norm.weight",
    "model.language_model.layers.43.self_attn.q_norm.weight",
    "model.language_model.layers.44.input_layernorm.weight",
    "model.language_model.layers.44.post_attention_layernorm.weight",
    "model.language_model.layers.44.self_attn.k_norm.weight",
    "model.language_model.layers.44.self_attn.q_norm.weight",
    "model.language_model.layers.45.input_layernorm.weight",
    "model.language_model.layers.45.post_attention_layernorm.weight",
    "model.language_model.layers.45.self_attn.k_norm.weight",
    "model.language_model.layers.45.self_attn.q_norm.weight",
    "model.language_model.layers.46.input_layernorm.weight",
    "model.language_model.layers.46.post_attention_layernorm.weight",
    "model.language_model.layers.46.self_attn.k_norm.weight",
    "model.language_model.layers.46.self_attn.q_norm.weight",
    "model.language_model.layers.47.input_layernorm.weight",
    "model.language_model.layers.47.post_attention_layernorm.weight",
    "model.language_model.layers.47.self_attn.k_norm.weight",
    "model.language_model.layers.47.self_attn.q_norm.weight",
    "model.language_model.layers.48.input_layernorm.weight",
    "model.language_model.layers.48.post_attention_layernorm.weight",
    "model.language_model.layers.48.self_attn.k_norm.weight",
    "model.language_model.layers.48.self_attn.q_norm.weight",
    "model.language_model.layers.49.input_layernorm.weight",
    "model.language_model.layers.49.post_attention_layernorm.weight",
    "model.language_model.layers.49.self_attn.k_norm.weight",
    "model.language_model.layers.49.self_attn.q_norm.weight",
    "model.language_model.layers.50.input_layernorm.weight",
    "model.language_model.layers.50.post_attention_layernorm.weight",
    "model.language_model.layers.50.self_attn.k_norm.weight",
    "model.language_model.layers.50.self_attn.q_norm.weight",
    "model.language_model.layers.51.input_layernorm.weight",
    "model.language_model.layers.51.post_attention_layernorm.weight",
    "model.language_model.layers.51.self_attn.k_norm.weight",
    "model.language_model.layers.51.self_attn.q_norm.weight",
    "model.language_model.layers.52.input_layernorm.weight",
    "model.language_model.layers.52.post_attention_layernorm.weight",
    "model.language_model.layers.52.self_attn.k_norm.weight",
    "model.language_model.layers.52.self_attn.q_norm.weight",
    "model.language_model.layers.53.input_layernorm.weight",
    "model.language_model.layers.53.post_attention_layernorm.weight",
    "model.language_model.layers.53.self_attn.k_norm.weight",
    "model.language_model.layers.53.self_attn.q_norm.weight",
    "model.language_model.layers.54.input_layernorm.weight",
    "model.language_model.layers.54.post_attention_layernorm.weight",
    "model.language_model.layers.54.self_attn.k_norm.weight",
    "model.language_model.layers.54.self_attn.q_norm.weight",
    "model.language_model.layers.55.input_layernorm.weight",
    "model.language_model.layers.55.post_attention_layernorm.weight",
    "model.language_model.layers.55.self_attn.k_norm.weight",
    "model.language_model.layers.55.self_attn.q_norm.weight",
    "model.language_model.layers.56.input_layernorm.weight",
    "model.language_model.layers.56.post_attention_layernorm.weight",
    "model.language_model.layers.56.self_attn.k_norm.weight",
    "model.language_model.layers.56.self_attn.q_norm.weight",
    "model.language_model.layers.57.input_layernorm.weight",
    "model.language_model.layers.57.post_attention_layernorm.weight",
    "model.language_model.layers.57.self_attn.k_norm.weight",
    "model.language_model.layers.57.self_attn.q_norm.weight",
    "model.language_model.layers.58.input_layernorm.weight",
    "model.language_model.layers.58.post_attention_layernorm.weight",
    "model.language_model.layers.58.self_attn.k_norm.weight",
    "model.language_model.layers.58.self_attn.q_norm.weight",
    "model.language_model.layers.59.input_layernorm.weight",
    "model.language_model.layers.59.post_attention_layernorm.weight",
    "model.language_model.layers.59.self_attn.k_norm.weight",
    "model.language_model.layers.59.self_attn.q_norm.weight",
    "model.language_model.layers.60.input_layernorm.weight",
    "model.language_model.layers.60.post_attention_layernorm.weight",
    "model.language_model.layers.60.self_attn.k_norm.weight",
    "model.language_model.layers.60.self_attn.q_norm.weight",
    "model.language_model.layers.61.input_layernorm.weight",
    "model.language_model.layers.61.post_attention_layernorm.weight",
    "model.language_model.layers.61.self_attn.k_norm.weight",
    "model.language_model.layers.61.self_attn.q_norm.weight",
    "model.language_model.layers.62.input_layernorm.weight",
    "model.language_model.layers.62.post_attention_layernorm.weight",
    "model.language_model.layers.62.self_attn.k_norm.weight",
    "model.language_model.layers.62.self_attn.q_norm.weight",
    "model.language_model.layers.63.input_layernorm.weight",
    "model.language_model.layers.63.post_attention_layernorm.weight",
    "model.language_model.layers.63.self_attn.k_norm.weight",
    "model.language_model.layers.63.self_attn.q_norm.weight",
    "lm_head.weight",
    "model.language_model.norm.weight",
    "model.visual.blocks.0.norm1.weight",
    "model.visual.blocks.0.norm2.weight",
    "model.visual.blocks.1.norm1.weight",
    "model.visual.blocks.1.norm2.weight",
    "model.visual.blocks.10.norm1.weight",
    "model.visual.blocks.10.norm2.weight",
    "model.visual.blocks.11.norm1.weight",
    "model.visual.blocks.11.norm2.weight",
    "model.visual.blocks.12.norm1.weight",
    "model.visual.blocks.12.norm2.weight",
    "model.visual.blocks.13.norm1.weight",
    "model.visual.blocks.13.norm2.weight",
    "model.visual.blocks.14.norm1.weight",
    "model.visual.blocks.14.norm2.weight",
    "model.visual.blocks.15.norm1.weight",
    "model.visual.blocks.15.norm2.weight",
    "model.visual.blocks.16.norm1.weight",
    "model.visual.blocks.16.norm2.weight",
    "model.visual.blocks.17.norm1.weight",
    "model.visual.blocks.17.norm2.weight",
    "model.visual.blocks.18.norm1.weight",
    "model.visual.blocks.18.norm2.weight",
    "model.visual.blocks.19.norm1.weight",
    "model.visual.blocks.19.norm2.weight",
    "model.visual.blocks.2.norm1.weight",
    "model.visual.blocks.2.norm2.weight",
    "model.visual.blocks.20.norm1.weight",
    "model.visual.blocks.20.norm2.weight",
    "model.visual.blocks.21.norm1.weight",
    "model.visual.blocks.21.norm2.weight",
    "model.visual.blocks.22.norm1.weight",
    "model.visual.blocks.22.norm2.weight",
    "model.visual.blocks.23.norm1.weight",
    "model.visual.blocks.23.norm2.weight",
    "model.visual.blocks.24.norm1.weight",
    "model.visual.blocks.24.norm2.weight",
    "model.visual.blocks.25.norm1.weight",
    "model.visual.blocks.25.norm2.weight",
    "model.visual.blocks.26.norm1.weight",
    "model.visual.blocks.26.norm2.weight",
    "model.visual.blocks.3.norm1.weight",
    "model.visual.blocks.3.norm2.weight",
    "model.visual.blocks.4.norm1.weight",
    "model.visual.blocks.4.norm2.weight",
    "model.visual.blocks.5.norm1.weight",
    "model.visual.blocks.5.norm2.weight",
    "model.visual.blocks.6.norm1.weight",
    "model.visual.blocks.6.norm2.weight",
    "model.visual.blocks.7.norm1.weight",
    "model.visual.blocks.7.norm2.weight",
    "model.visual.blocks.8.norm1.weight",
    "model.visual.blocks.8.norm2.weight",
    "model.visual.blocks.9.norm1.weight",
    "model.visual.blocks.9.norm2.weight",
    "model.visual.deepstack_merger_list.0.norm.weight",
    "model.visual.deepstack_merger_list.1.norm.weight",
    "model.visual.deepstack_merger_list.2.norm.weight",
    "model.visual.merger.norm.weight",
    "model.visual.patch_embed.proj.weight",
    "model.visual.pos_embed.weight"
  ],
  "modules_to_not_use_matmul": [],
  "non_blocking": false,
  "quant_conv": false,
  "quant_embedding": false,
  "quant_method": "sdnq",
  "quantization_device": "cpu",
  "quantized_matmul_dtype": null,
  "return_device": "cpu",
  "sdnq_version": "0.2.5+ai.manager.5",
  "svd_rank": 32,
  "svd_steps": 8,
  "use_codebook": false,
  "use_dynamic_quantization": false,
  "use_grad_ckpt": true,
  "use_hadamard": false,
  "use_quantized_matmul": false,
  "use_quantized_matmul_conv": false,
  "use_static_quantization": true,
  "use_stochastic_rounding": false,
  "use_svd": false,
  "weights_dtype": "uint4"
}