leninangelov/basic-chat-model
052
1{
2 "architectures": [
3 "T5ForConditionalGeneration"
4 ],
5 "classifier_dropout": 0.0,
6 "d_ff": 1792,
7 "d_kv": 64,
8 "d_model": 384,
9 "decoder_start_token_id": 954,
10 "dense_act_fn": "relu",
11 "dropout_rate": 0.1,
12 "eos_token_id": 1,
13 "feed_forward_proj": "relu",
14 "initializer_factor": 1.0,
15 "is_encoder_decoder": true,
16 "is_gated_act": false,
17 "layer_norm_epsilon": 1e-05,
18 "model_type": "t5",
19 "num_decoder_layers": 4,
20 "num_heads": 12,
21 "num_layers": 4,
22 "pad_token_id": 954,
23 "relative_attention_max_distance": 128,
24 "relative_attention_num_buckets": 32,
25 "torch_dtype": "float32",
26 "transformers_version": "4.45.2",
27 "use_cache": true,
28 "vocab_size": 14000
29}
30 