bigscience/bigscience-small-testing
49.8k
1{2 "_name_or_path": "/home/younes/Desktop/Work/data/megatron-debug/",3 "apply_residual_connection_post_layernorm": false,4 "architectures": [5 "BloomModel"6 ],7 "attention_dropout": 0.1,8 "bias_dropout_fusion": true,9 "bos_token_id": 0,10 "dtype": "bfloat16",11 "eos_token_id": 0,12 "hidden_dropout": 0.1,13 "hidden_size": 64,14 "initializer_range": 0.02,15 "layer_norm_epsilon": 1e-05,16 "masked_softmax_fusion": true,17 "model_type": "bloom",18 "n_head": 8,19 "n_inner": null,20 "n_layer": 2,21 "pretraining_tp": 2,22 "seq_length": 20,23 "skip_bias_add": true,24 "torch_dtype": "bfloat16",25 "transformers_version": "4.18.0",26 "use_cache": false,27 "vocab_size": 250880,28 "slow_but_exact": true29}30 