hf-tiny-model-private/tiny-random-XGLMForCausalLM
0107
1{2 "_name_or_path": "tiny_models/xglm/XGLMForCausalLM",3 "activation_dropout": 0.1,4 "activation_function": "gelu",5 "architectures": [6 "XGLMForCausalLM"7 ],8 "attention_dropout": 0.1,9 "attention_heads": 4,10 "bos_token_id": 0,11 "d_model": 32,12 "decoder_start_token_id": 2,13 "dropout": 0.1,14 "eos_token_id": 2,15 "ffn_dim": 37,16 "gradient_checkpointing": false,17 "init_std": 0.02,18 "initializer_range": 0.02,19 "is_decoder": true,20 "layerdrop": 0.0,21 "max_position_embeddings": 512,22 "model_type": "xglm",23 "num_layers": 5,24 "pad_token_id": 1,25 "scale_embedding": true,26 "torch_dtype": "float32",27 "transformers_version": "4.28.0.dev0",28 "use_cache": true,29 "vocab_size": 25600830}31 