Push model to huggingface

1_Pooling/config.json

@@ -1,5 +1,5 @@
 {
-  "word_embedding_dimension": 384,
+  "word_embedding_dimension": 768,
   "pooling_mode_cls_token": false,
   "pooling_mode_mean_tokens": true,
   "pooling_mode_max_tokens": false,

README.md

@@ -11,13 +11,14 @@ datasets:
 - Salesforce/dialogstudio
 pipeline_tag: sentence-similarity
 base_model:
-- google-bert/bert-base-uncased
+- aws-ai/dse-bert-base
 ---
 
 
-# Dialog2Flow joint target (BERT-base)
+# Dialog2Flow single target (DSE-base)
 
-This is the original **D2F$_{joint}$** model introduced in the paper ["Dialog2Flow: Pre-training Soft-Contrastive Action-Driven Sentence Embeddings for Automatic Dialog Flow Extraction"](https://publications.idiap.ch/attachments/papers/2024/Burdisso_EMNLP2024_2024.pdf) published in the EMNLP 2024 main conference.
+This a variation of the **D2F$_{single}$** model introduced in the paper ["Dialog2Flow: Pre-training Soft-Contrastive Action-Driven Sentence Embeddings for Automatic Dialog Flow Extraction"](https://publications.idiap.ch/attachments/papers/2024/Burdisso_EMNLP2024_2024.pdf) published in the EMNLP 2024 main conference.
+This version uses DSE-base as the backbone model which yields to an increase in performance as compared to the vanilla version using BERT-base as the backbone (results reported in Appendix C).
 
 Implementation-wise, this is a [sentence-transformers](https://www.SBERT.net) model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or search.
 
@@ -37,7 +38,7 @@ Then you can use the model like this:
 from sentence_transformers import SentenceTransformer
 sentences = ["your phone please", "okay may i have your telephone number please"]
 
-model = SentenceTransformer('sergioburdisso/dialog2flow-joint-bert-base')
+model = SentenceTransformer('sergioburdisso/dialog2flow-single-dse-base')
 embeddings = model.encode(sentences)
 print(embeddings)
 ```
@@ -63,8 +64,8 @@ def mean_pooling(model_output, attention_mask):
 sentences = ['your phone please', 'okay may i have your telephone number please']
 
 # Load model from HuggingFace Hub
-tokenizer = AutoTokenizer.from_pretrained('sergioburdisso/dialog2flow-joint-bert-base')
-model = AutoModel.from_pretrained('sergioburdisso/dialog2flow-joint-bert-base')
+tokenizer = AutoTokenizer.from_pretrained('sergioburdisso/dialog2flow-single-dse-base')
+model = AutoModel.from_pretrained('sergioburdisso/dialog2flow-single-dse-base')
 
 # Tokenize sentences
 encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
@@ -153,4 +154,4 @@ SentenceTransformer(
 ## License
 
 Copyright (c) 2024 [Idiap Research Institute](https://www.idiap.ch/).
-MIT License.
+MIT License.

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "sentence-transformers/all-MiniLM-L6-v2",
+  "_name_or_path": "/idiap/temp/sburdisso/repos/jsalt/keya-dialog/outputs/tod_das+slots/bert-base-uncased/soft-labels/label_multi-qa-mpnet-base-dot-v1_t0.35/msl64_pm-mean/ch-on_t0.05/lr3e-06_bs64_e15/best_model_metric_0",
   "architectures": [
     "BertModel"
   ],
@@ -8,14 +8,14 @@
   "gradient_checkpointing": false,
   "hidden_act": "gelu",
   "hidden_dropout_prob": 0.1,
-  "hidden_size": 384,
+  "hidden_size": 768,
   "initializer_range": 0.02,
-  "intermediate_size": 1536,
+  "intermediate_size": 3072,
   "layer_norm_eps": 1e-12,
   "max_position_embeddings": 512,
   "model_type": "bert",
   "num_attention_heads": 12,
-  "num_hidden_layers": 6,
+  "num_hidden_layers": 12,
   "pad_token_id": 0,
   "position_embedding_type": "absolute",
   "torch_dtype": "float32",

config_sentence_transformers.json

@@ -1,7 +1,7 @@
 {
   "__version__": {
-    "sentence_transformers": "2.0.0",
-    "transformers": "4.6.1",
-    "pytorch": "1.8.1"
+    "sentence_transformers": "2.2.2",
+    "transformers": "4.30.2",
+    "pytorch": "2.0.1"
   }
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1377e9af0ca0b016a9f2aa584d6fc71ab3ea6804fae21ef9fb1416e2944057ac
-size 90864192
+oid sha256:133a5d2947ff9797b9bddeef74fcc957f7485fc0d219e59362e8489e9a4c3b76
+size 437951328

modules.json

@@ -10,11 +10,5 @@
     "name": "1",
     "path": "1_Pooling",
     "type": "sentence_transformers.models.Pooling"
-  },
-  {
-    "idx": 2,
-    "name": "2",
-    "path": "2_Normalize",
-    "type": "sentence_transformers.models.Normalize"
   }
 ]

sentence_bert_config.json

@@ -1,4 +1,4 @@
 {
-  "max_seq_length": 256,
+  "max_seq_length": 64,
   "do_lower_case": false
 }

tokenizer.json

@@ -2,14 +2,12 @@
   "version": "1.0",
   "truncation": {
     "direction": "Right",
-    "max_length": 128,
+    "max_length": 64,
     "strategy": "LongestFirst",
     "stride": 0
   },
   "padding": {
-    "strategy": {
-      "Fixed": 128
-    },
+    "strategy": "BatchLongest",
     "direction": "Right",
     "pad_to_multiple_of": null,
     "pad_id": 0,

tokenizer_config.json

@@ -43,12 +43,10 @@
   },
   "clean_up_tokenization_spaces": true,
   "cls_token": "[CLS]",
-  "do_basic_tokenize": true,
   "do_lower_case": true,
   "mask_token": "[MASK]",
-  "max_length": 128,
+  "max_length": 64,
   "model_max_length": 512,
-  "never_split": null,
   "pad_to_multiple_of": null,
   "pad_token": "[PAD]",
   "pad_token_type_id": 0,