config.json

{
  "_name_or_path": "manoskary/NotaGenX-Quantized",
  "architectures": [
    "NotaGenLMHeadModel"
  ],
  "model_type": "notagen",
  "quantization_config": {
    "quantization_method": "pytorch_dynamic_int8",
    "target_modules": [
      "linear",
      "embedding"
    ],
    "preserved_accuracy": 0.98
  },
  "patch_config": {
    "num_hidden_layers": 20,
    "max_length": 1024,
    "max_position_embeddings": 1024,
    "n_embd": 1280,
    "num_attention_heads": 20,
    "vocab_size": 1
  },
  "decoder_config": {
    "num_hidden_layers": 6,
    "max_length": 17,
    "max_position_embeddings": 17,
    "hidden_size": 1280,
    "num_attention_heads": 20,
    "vocab_size": 128
  }
}