hf-tiny-model-private/tiny-random-ReformerModel

1{2  "architectures": [3    "ReformerModel"4  ],5  "attention_head_size": 64,6  "attn_layers": [7    "local",8    "local",9    "local",10    "local"11  ],12  "axial_norm_std": 1.0,13  "axial_pos_embds": true,14  "axial_pos_embds_dim": [15    16,16    1617  ],18  "axial_pos_shape": [19    4,20    2521  ],22  "chunk_size_lm_head": 0,23  "classifier_dropout": null,24  "eos_token_id": 2,25  "feed_forward_size": 32,26  "hash_seed": 0,27  "hidden_act": "gelu",28  "hidden_dropout_prob": 0.1,29  "hidden_size": 32,30  "initializer_range": 0.02,31  "layer_norm_eps": 1e-12,32  "local_attention_probs_dropout_prob": 0.1,33  "local_attn_chunk_length": 4,34  "local_num_chunks_after": 0,35  "local_num_chunks_before": 1,36  "lsh_attention_probs_dropout_prob": null,37  "lsh_attn_chunk_length": null,38  "lsh_num_chunks_after": null,39  "lsh_num_chunks_before": null,40  "max_position_embeddings": 100,41  "model_type": "reformer",42  "num_attention_heads": 2,43  "num_buckets": null,44  "num_hashes": 1,45  "num_hidden_layers": 4,46  "pad_token_id": 0,47  "tie_word_embeddings": false,48  "torch_dtype": "float32",49  "transformers_version": "4.28.0.dev0",50  "use_cache": true,51  "vocab_size": 32052}53