Upload DPLM2ForMaskedLM
Browse files- config.json +9 -8
- model.safetensors +3 -0
config.json
CHANGED
|
@@ -1,20 +1,21 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "airkingbd/dplm_3b",
|
| 3 |
"aa_type": 1,
|
|
|
|
| 4 |
"architectures": [
|
| 5 |
"DPLM2ForMaskedLM"
|
| 6 |
],
|
| 7 |
"attention_probs_dropout_prob": 0.0,
|
| 8 |
"attn_backend": "sdpa",
|
| 9 |
"auto_map": {
|
| 10 |
-
"AutoConfig": "
|
| 11 |
-
"AutoModel": "
|
| 12 |
-
"AutoModelForMaskedLM": "
|
| 13 |
-
"AutoModelForSequenceClassification": "
|
| 14 |
-
"AutoModelForTokenClassification": "
|
| 15 |
},
|
| 16 |
"classifier_dropout": null,
|
| 17 |
"dplm_type": "dplm_esm",
|
|
|
|
| 18 |
"emb_layer_norm_before": false,
|
| 19 |
"esmfold_config": null,
|
| 20 |
"hidden_act": "gelu",
|
|
@@ -22,6 +23,7 @@
|
|
| 22 |
"hidden_size": 2560,
|
| 23 |
"initializer_range": 0.02,
|
| 24 |
"intermediate_size": 10240,
|
|
|
|
| 25 |
"is_folding_model": false,
|
| 26 |
"layer_norm_eps": 1e-05,
|
| 27 |
"mask_token_id": 32,
|
|
@@ -35,8 +37,7 @@
|
|
| 35 |
"struct_type": 0,
|
| 36 |
"tie_word_embeddings": false,
|
| 37 |
"token_dropout": true,
|
| 38 |
-
"
|
| 39 |
-
"transformers_version": "4.48.1",
|
| 40 |
"use_cache": true,
|
| 41 |
"vocab_list": null,
|
| 42 |
"vocab_size": 8229
|
|
|
|
| 1 |
{
|
|
|
|
| 2 |
"aa_type": 1,
|
| 3 |
+
"add_cross_attention": false,
|
| 4 |
"architectures": [
|
| 5 |
"DPLM2ForMaskedLM"
|
| 6 |
],
|
| 7 |
"attention_probs_dropout_prob": 0.0,
|
| 8 |
"attn_backend": "sdpa",
|
| 9 |
"auto_map": {
|
| 10 |
+
"AutoConfig": "modeling_dplm2.DPLM2Config",
|
| 11 |
+
"AutoModel": "modeling_dplm2.DPLM2Model",
|
| 12 |
+
"AutoModelForMaskedLM": "modeling_dplm2.DPLM2ForMaskedLM",
|
| 13 |
+
"AutoModelForSequenceClassification": "modeling_dplm2.DPLM2ForSequenceClassification",
|
| 14 |
+
"AutoModelForTokenClassification": "modeling_dplm2.DPLM2ForTokenClassification"
|
| 15 |
},
|
| 16 |
"classifier_dropout": null,
|
| 17 |
"dplm_type": "dplm_esm",
|
| 18 |
+
"dtype": "float32",
|
| 19 |
"emb_layer_norm_before": false,
|
| 20 |
"esmfold_config": null,
|
| 21 |
"hidden_act": "gelu",
|
|
|
|
| 23 |
"hidden_size": 2560,
|
| 24 |
"initializer_range": 0.02,
|
| 25 |
"intermediate_size": 10240,
|
| 26 |
+
"is_decoder": false,
|
| 27 |
"is_folding_model": false,
|
| 28 |
"layer_norm_eps": 1e-05,
|
| 29 |
"mask_token_id": 32,
|
|
|
|
| 37 |
"struct_type": 0,
|
| 38 |
"tie_word_embeddings": false,
|
| 39 |
"token_dropout": true,
|
| 40 |
+
"transformers_version": "5.2.0",
|
|
|
|
| 41 |
"use_cache": true,
|
| 42 |
"vocab_list": null,
|
| 43 |
"vocab_size": 8229
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77767a93408a01e8485ae27b008712aa8dbd6fa4221f97879a2703d45489e9be
|
| 3 |
+
size 11524320476
|