add config file
#2
by
dmarpina
- opened
- config.json +47 -0
config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": ["T5ForConditionalGeneration"],
|
| 3 |
+
"classifier_dropout": 0.0,
|
| 4 |
+
"d_ff": 1024,
|
| 5 |
+
"d_kv": 64,
|
| 6 |
+
"d_model": 256,
|
| 7 |
+
"decoder_start_token_id": 0,
|
| 8 |
+
"dense_act_fn": "relu",
|
| 9 |
+
"dropout_rate": 0.1,
|
| 10 |
+
"eos_token_id": 1,
|
| 11 |
+
"feed_forward_proj": "relu",
|
| 12 |
+
"initializer_factor": 0.05,
|
| 13 |
+
"is_encoder_decoder": true,
|
| 14 |
+
"is_gated_act": false,
|
| 15 |
+
"layer_norm_epsilon": 1e-6,
|
| 16 |
+
"model_type": "t5",
|
| 17 |
+
"n_positions": 512,
|
| 18 |
+
"num_decoder_layers": 4,
|
| 19 |
+
"num_heads": 4,
|
| 20 |
+
"num_layers": 4,
|
| 21 |
+
"pad_token_id": 0,
|
| 22 |
+
"relative_attention_max_distance": 128,
|
| 23 |
+
"relative_attention_num_buckets": 32,
|
| 24 |
+
"torch_dtype": "float32",
|
| 25 |
+
"transformers_version": "4.37.2",
|
| 26 |
+
"use_cache": true,
|
| 27 |
+
"vocab_size": 4096,
|
| 28 |
+
"chronos_config": {
|
| 29 |
+
"tokenizer_class": "MeanScaleUniformBins",
|
| 30 |
+
"tokenizer_kwargs": {
|
| 31 |
+
"low_limit": -15.0,
|
| 32 |
+
"high_limit": 15.0
|
| 33 |
+
},
|
| 34 |
+
"n_tokens": 4096,
|
| 35 |
+
"n_special_tokens": 2,
|
| 36 |
+
"pad_token_id": 0,
|
| 37 |
+
"eos_token_id": 1,
|
| 38 |
+
"use_eos_token": true,
|
| 39 |
+
"model_type": "seq2seq",
|
| 40 |
+
"context_length": 512,
|
| 41 |
+
"prediction_length": 64,
|
| 42 |
+
"num_samples": 20,
|
| 43 |
+
"temperature": 1.0,
|
| 44 |
+
"top_k": 50,
|
| 45 |
+
"top_p": 1.0
|
| 46 |
+
}
|
| 47 |
+
}
|