CreitinGameplays
commited on
Commit
•
6ddc08d
1
Parent(s):
f6df113
Upload config.json
Browse files- config.json +26 -0
config.json
ADDED
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "TurkuNLP/gpt3-finnish-xl",
|
3 |
+
"apply_residual_connection_post_layernorm": false,
|
4 |
+
"architectures": [
|
5 |
+
"BloomForCausalLM"
|
6 |
+
],
|
7 |
+
"attention_dropout": 0.0,
|
8 |
+
"attention_softmax_in_fp32": true,
|
9 |
+
"bos_token_id": 1,
|
10 |
+
"eos_token_id": 2,
|
11 |
+
"hidden_dropout": 0.0,
|
12 |
+
"hidden_size": 2064,
|
13 |
+
"initializer_range": 0.02,
|
14 |
+
"layer_norm_epsilon": 1e-05,
|
15 |
+
"masked_softmax_fusion": true,
|
16 |
+
"model_type": "bloom",
|
17 |
+
"n_head": 24,
|
18 |
+
"n_layer": 24,
|
19 |
+
"pad_token_id": 3,
|
20 |
+
"pretraining_tp": 1,
|
21 |
+
"slow_but_exact": false,
|
22 |
+
"torch_dtype": "float16",
|
23 |
+
"transformers_version": "4.38.2",
|
24 |
+
"use_cache": true,
|
25 |
+
"vocab_size": 131072
|
26 |
+
}
|