config.json

{
  "alibi_mode": "symetric",
  "architectures": [
    "FlashT5ForConditionalGeneration"
  ],
  "attention_dropout_rate": 0.0,
  "attention_scale": 1.0,
  "attention_type": "ref",
  "auto_map": {
    "AutoConfig": "configuration_flash_t5.FlashT5Config",
    "AutoModel": "modeling_flash_t5.FlashT5ForConditionalGeneration",
    "AutoModelForQuestionAnswering": "custom_heads_flash_t5.FlashT5ForQuestionAnswering",
    "AutoModelForSeq2SeqLM": "modeling_flash_t5.FlashT5ForConditionalGeneration",
    "AutoModelForSequenceClassification": "custom_heads_flash_t5.FlashT5ForSequenceClassification",
    "AutoModelForTokenClassification": "custom_heads_flash_t5.FlashT5ForTokenClassification"
  },
  "classifier_dropout": 0.0,
  "d_ff": 2816,
  "d_kv": 64,
  "d_model": 1024,
  "decoder_start_token_id": 0,
  "dense_act_fn": "relu",
  "dropout_rate": 0.0,
  "eos_token_id": 1,
  "feed_forward_proj": "relu",
  "fire_mlp_width": 32,
  "initializer_factor": 1.0,
  "is_encoder_decoder": false,
  "is_gated_act": false,
  "label_smoothing": 0.1,
  "layer_norm_epsilon": 1e-06,
  "max_sequence_length": 1024,
  "model_type": "flash_t5",
  "num_decoder_layers": 24,
  "num_heads": 16,
  "num_layers": 24,
  "pad_token_id": 0,
  "position_encoding_type": "t5",
  "relative_attention_max_distance": 128,
  "relative_attention_num_buckets": 32,
  "rotary_base": 10000,
  "rotary_emb_fraction": 1.0,
  "rotary_interleaved": false,
  "rotary_scale_base": null,
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.39.3",
  "use_cache": true,
  "use_flash_attention": "triton",
  "use_full_bias_size": false,
  "use_gelu_act": true,
  "use_glu_mlp": true,
  "use_masking": false,
  "use_randomized_position_encoding": false,
  "use_triton_crossentropy": false,
  "use_triton_gated_mlp": false,
  "use_triton_layernorm": false,
  "vocab_size": 32128,
  "z_loss": 0.0001
}