bumblebee-testing/tiny-random-T5ForConditionalGeneration-tie_word_embeddings-False
087
1{2 "_name_or_path": "hf-internal-testing/tiny-random-T5ForConditionalGeneration",3 "architectures": [4 "T5ForConditionalGeneration"5 ],6 "bos_token_id": 0,7 "classifier_dropout": 0.0,8 "d_ff": 37,9 "d_kv": 8,10 "d_model": 32,11 "decoder_start_token_id": 0,12 "dense_act_fn": "relu",13 "dropout_rate": 0.1,14 "eos_token_id": 1,15 "feed_forward_proj": "relu",16 "initializer_factor": 0.002,17 "is_encoder_decoder": true,18 "is_gated_act": false,19 "layer_norm_epsilon": 1e-06,20 "model_type": "t5",21 "num_decoder_layers": 5,22 "num_heads": 4,23 "num_layers": 5,24 "pad_token_id": 0,25 "relative_attention_max_distance": 128,26 "relative_attention_num_buckets": 8,27 "tie_word_embeddings": false,28 "torch_dtype": "float32",29 "transformers_version": "4.36.0.dev0",30 "use_cache": true,31 "vocab_size": 3210032}33 