sieg2011/codet5-base-sql-create-context
016
1{
2 "architectures": [
3 "T5ForConditionalGeneration"
4 ],
5 "bos_token_id": 1,
6 "classifier_dropout": 0.0,
7 "d_ff": 3072,
8 "d_kv": 64,
9 "d_model": 768,
10 "decoder_start_token_id": 0,
11 "dense_act_fn": "relu",
12 "dropout_rate": 0.1,
13 "eos_token_id": 2,
14 "feed_forward_proj": "relu",
15 "gradient_checkpointing": false,
16 "id2label": {
17 "0": "LABEL_0"
18 },
19 "initializer_factor": 1.0,
20 "is_encoder_decoder": true,
21 "is_gated_act": false,
22 "label2id": {
23 "LABEL_0": 0
24 },
25 "layer_norm_epsilon": 1e-06,
26 "model_type": "t5",
27 "n_positions": 512,
28 "num_decoder_layers": 12,
29 "num_heads": 12,
30 "num_layers": 12,
31 "output_past": true,
32 "pad_token_id": 0,
33 "relative_attention_max_distance": 128,
34 "relative_attention_num_buckets": 32,
35 "task_specific_params": {
36 "summarization": {
37 "early_stopping": true,
38 "length_penalty": 2.0,
39 "max_length": 200,
40 "min_length": 30,
41 "no_repeat_ngram_size": 3,
42 "num_beams": 4,
43 "prefix": "summarize: "
44 },
45 "translation_en_to_de": {
46 "early_stopping": true,
47 "max_length": 300,
48 "num_beams": 4,
49 "prefix": "translate English to German: "
50 },
51 "translation_en_to_fr": {
52 "early_stopping": true,
53 "max_length": 300,
54 "num_beams": 4,
55 "prefix": "translate English to French: "
56 },
57 "translation_en_to_ro": {
58 "early_stopping": true,
59 "max_length": 300,
60 "num_beams": 4,
61 "prefix": "translate English to Romanian: "
62 }
63 },
64 "torch_dtype": "float32",
65 "transformers_version": "4.50.3",
66 "use_cache": true,
67 "vocab_size": 32100
68}
69 