Bharatdeep-H/pq_cache_3

1{2  "_name_or_path": "jinaai/jina-embedding-b-en-v1",3  "architectures": [4    "T5EncoderModel"5  ],6  "auto_map": {7    "AutoModel": "jinaai/jina-embedding-b-en-v1--modeling_t5.T5EncoderModel"8  },9  "classifier_dropout": 0.0,10  "d_ff": 3072,11  "d_kv": 64,12  "d_model": 768,13  "decoder_start_token_id": 0,14  "dense_act_fn": "relu",15  "dropout_rate": 0.1,16  "eos_token_id": 1,17  "feed_forward_proj": "relu",18  "initializer_factor": 1.0,19  "is_encoder_decoder": true,20  "is_gated_act": false,21  "layer_norm_epsilon": 1e-06,22  "model_type": "t5",23  "n_positions": 512,24  "num_decoder_layers": 12,25  "num_heads": 12,26  "num_layers": 12,27  "output_past": true,28  "pad_token_id": 0,29  "relative_attention_max_distance": 128,30  "relative_attention_num_buckets": 32,31  "task_specific_params": {32    "summarization": {33      "early_stopping": true,34      "length_penalty": 2.0,35      "max_length": 200,36      "min_length": 30,37      "no_repeat_ngram_size": 3,38      "num_beams": 4,39      "prefix": "summarize: "40    },41    "translation_en_to_de": {42      "early_stopping": true,43      "max_length": 300,44      "num_beams": 4,45      "prefix": "translate English to German: "46    },47    "translation_en_to_fr": {48      "early_stopping": true,49      "max_length": 300,50      "num_beams": 4,51      "prefix": "translate English to French: "52    },53    "translation_en_to_ro": {54      "early_stopping": true,55      "max_length": 300,56      "num_beams": 4,57      "prefix": "translate English to Romanian: "58    }59  },60  "torch_dtype": "float32",61  "transformers_version": "4.49.0",62  "use_cache": true,63  "vocab_size": 3212864}65