Commit
·
df17b75
1
Parent(s):
005f82d
Upload 4 files
Browse files- config.json +6 -6
config.json
CHANGED
|
@@ -2,9 +2,9 @@
|
|
| 2 |
"architectures": [
|
| 3 |
"T5ForConditionalGeneration"
|
| 4 |
],
|
| 5 |
-
"d_ff":
|
| 6 |
"d_kv": 64,
|
| 7 |
-
"d_model":
|
| 8 |
"decoder_start_token_id": 0,
|
| 9 |
"dropout_rate": 0.1,
|
| 10 |
"eos_token_id": 1,
|
|
@@ -13,8 +13,8 @@
|
|
| 13 |
"layer_norm_epsilon": 1e-06,
|
| 14 |
"model_type": "t5",
|
| 15 |
"n_positions": 512,
|
| 16 |
-
"num_heads":
|
| 17 |
-
"num_layers":
|
| 18 |
"output_past": true,
|
| 19 |
"pad_token_id": 0,
|
| 20 |
"relative_attention_num_buckets": 32,
|
|
@@ -22,8 +22,8 @@
|
|
| 22 |
"summarization": {
|
| 23 |
"early_stopping": true,
|
| 24 |
"length_penalty": 2.0,
|
| 25 |
-
"max_length":
|
| 26 |
-
"min_length":
|
| 27 |
"no_repeat_ngram_size": 3,
|
| 28 |
"num_beams": 4,
|
| 29 |
"prefix": "summarize: "
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
"T5ForConditionalGeneration"
|
| 4 |
],
|
| 5 |
+
"d_ff": 2048,
|
| 6 |
"d_kv": 64,
|
| 7 |
+
"d_model": 512,
|
| 8 |
"decoder_start_token_id": 0,
|
| 9 |
"dropout_rate": 0.1,
|
| 10 |
"eos_token_id": 1,
|
|
|
|
| 13 |
"layer_norm_epsilon": 1e-06,
|
| 14 |
"model_type": "t5",
|
| 15 |
"n_positions": 512,
|
| 16 |
+
"num_heads": 8,
|
| 17 |
+
"num_layers": 6,
|
| 18 |
"output_past": true,
|
| 19 |
"pad_token_id": 0,
|
| 20 |
"relative_attention_num_buckets": 32,
|
|
|
|
| 22 |
"summarization": {
|
| 23 |
"early_stopping": true,
|
| 24 |
"length_penalty": 2.0,
|
| 25 |
+
"max_length": 5000,
|
| 26 |
+
"min_length": 2000,
|
| 27 |
"no_repeat_ngram_size": 3,
|
| 28 |
"num_beams": 4,
|
| 29 |
"prefix": "summarize: "
|