bumblebee-testing/tiny-random-PhiForSequenceClassification
040
1{2 "_name_or_path": "hf-internal-testing/tiny-random-PhiForSequenceClassification",3 "architectures": [4 "PhiForSequenceClassification"5 ],6 "attention_dropout": 0.0,7 "attention_probs_dropout_prob": 0.1,8 "bos_token_id": 0,9 "embd_pdrop": 0.0,10 "eos_token_id": 0,11 "hidden_act": "gelu",12 "hidden_dropout_prob": 0.1,13 "hidden_size": 32,14 "initializer_range": 0.02,15 "intermediate_size": 37,16 "layer_norm_eps": 1e-05,17 "max_position_embeddings": 512,18 "model_type": "phi",19 "num_attention_heads": 4,20 "num_hidden_layers": 2,21 "num_key_value_heads": 4,22 "pad_token_id": 0,23 "partial_rotary_factor": 0.5,24 "qk_layernorm": false,25 "resid_pdrop": 0.0,26 "rope_scaling": null,27 "rope_theta": 10000.0,28 "tie_word_embeddings": false,29 "torch_dtype": "float32",30 "transformers_version": "4.38.2",31 "type_vocab_size": 16,32 "use_cache": true,33 "vocab_size": 102434}35 