GiacomoSignorile commited on
Commit
bd6ab73
·
verified ·
1 Parent(s): 8e441ee

Initial upload of fine-tuned PatentSBERTa Specialist (eHealth)

Browse files
README.md CHANGED
@@ -5,38 +5,717 @@ tags:
5
  - feature-extraction
6
  - dense
7
  - generated_from_trainer
8
- - dataset_size:2633
9
  - loss:MultipleNegativesRankingLoss
10
- - dataset_size:7861
11
  base_model: AI-Growth-Lab/PatentSBERTa
12
  widget:
13
- - source_sentence: CONSORTIUM OF PROBIOTICS.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  sentences:
15
- - NANOSTRUCTURED FILTERING MEMBRANE FOR ANALYTICAL APPLICATIONS.
16
- - CONSORTIUM OF PROBIOTICS.
17
- - Real-time communications over Bluetooth Low Energy.
18
- - source_sentence: Device for the collection and analysis of a biological fluid.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19
  sentences:
20
- - Device for the collection and analysis of a biological fluid.
21
- - STED MICROSCOPY BASED ON SYNCHRONOUS DETECTION OF FLUORESCENCE EMISSION.
22
- - Robotic apparatus for minimally invasive surgery.
23
- - source_sentence: Method for obtaining a porous semiconductor.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
  sentences:
25
- - 'Racetrack Logic: smart in-memory computing.'
26
- - Method for obtaining a porous semiconductor.
27
- - Salts of benzimidazole compounds, their use and synthetic preparation.
28
- - source_sentence: Enthalpy exchangers with polymeric membrane.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
29
  sentences:
30
- - Closure system for engine connecting rods.
31
- - Enthalpy exchangers with polymeric membrane.
32
- - Soluble protein with high angiogenic activity.
33
- - source_sentence: FLEXIBLE CAPACITIVE KEYBOARD FOR TEXTILE USE.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
34
  sentences:
35
- - 'GreenValve III: System for spherical segment valves.'
36
- - FLEXIBLE CAPACITIVE KEYBOARD FOR TEXTILE USE.
37
- - Mandibular advancement device.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
38
  pipeline_tag: sentence-similarity
39
  library_name: sentence-transformers
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
40
  ---
41
 
42
  # SentenceTransformer based on AI-Growth-Lab/PatentSBERTa
@@ -88,9 +767,9 @@ from sentence_transformers import SentenceTransformer
88
  model = SentenceTransformer("sentence_transformers_model_id")
89
  # Run inference
90
  sentences = [
91
- 'FLEXIBLE CAPACITIVE KEYBOARD FOR TEXTILE USE.',
92
- 'FLEXIBLE CAPACITIVE KEYBOARD FOR TEXTILE USE.',
93
- 'GreenValve III: System for spherical segment valves.',
94
  ]
95
  embeddings = model.encode(sentences)
96
  print(embeddings.shape)
@@ -99,9 +778,9 @@ print(embeddings.shape)
99
  # Get the similarity scores for the embeddings
100
  similarities = model.similarity(embeddings, embeddings)
101
  print(similarities)
102
- # tensor([[1.0000, 1.0000, 0.3686],
103
- # [1.0000, 1.0000, 0.3686],
104
- # [0.3686, 0.3686, 1.0000]])
105
  ```
106
 
107
  <!--
@@ -128,6 +807,33 @@ You can finetune this model on your own dataset.
128
  *List how the model may foreseeably be misused and address what users ought not to do with the model.*
129
  -->
130
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
131
  <!--
132
  ## Bias, Risks and Limitations
133
 
@@ -146,19 +852,19 @@ You can finetune this model on your own dataset.
146
 
147
  #### Unnamed Dataset
148
 
149
- * Size: 7,861 training samples
150
  * Columns: <code>sentence_0</code> and <code>sentence_1</code>
151
  * Approximate statistics based on the first 1000 samples:
152
- | | sentence_0 | sentence_1 |
153
- |:--------|:----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|
154
- | type | string | string |
155
- | details | <ul><li>min: 4 tokens</li><li>mean: 12.32 tokens</li><li>max: 32 tokens</li></ul> | <ul><li>min: 8 tokens</li><li>mean: 12.01 tokens</li><li>max: 36 tokens</li></ul> |
156
  * Samples:
157
- | sentence_0 | sentence_1 |
158
- |:--------------------------------------------------------------------------------------------------------------|:----------------------------------------------------------------|
159
- | <code>Genetic diagnostic technique for SCA1-3,6,7</code> | <code>Applicant/Organization: UNIV DEGLI STUDI DI TORINO</code> |
160
- | <code>Method for detecting Macrophomina phaseolina</code> | <code>Technical Classification: C12Q</code> |
161
- | <code>System for controlled administration of a substance from a human -body-implanted infusion device</code> | <code>Applicant/Organization: Stefanini Cesare</code> |
162
  * Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
163
  ```json
164
  {
@@ -171,105 +877,107 @@ You can finetune this model on your own dataset.
171
  ### Training Hyperparameters
172
  #### Non-Default Hyperparameters
173
 
 
174
  - `per_device_train_batch_size`: 12
175
  - `per_device_eval_batch_size`: 12
 
176
  - `multi_dataset_batch_sampler`: round_robin
177
 
178
  #### All Hyperparameters
179
  <details><summary>Click to expand</summary>
180
 
 
 
 
181
  - `per_device_train_batch_size`: 12
182
- - `num_train_epochs`: 3
183
- - `max_steps`: -1
 
 
184
  - `learning_rate`: 5e-05
185
- - `lr_scheduler_type`: linear
186
- - `lr_scheduler_kwargs`: None
187
- - `warmup_steps`: 0
188
- - `optim`: adamw_torch_fused
189
- - `optim_args`: None
190
  - `weight_decay`: 0.0
191
  - `adam_beta1`: 0.9
192
  - `adam_beta2`: 0.999
193
  - `adam_epsilon`: 1e-08
194
- - `optim_target_modules`: None
195
- - `gradient_accumulation_steps`: 1
196
- - `average_tokens_across_devices`: True
197
  - `max_grad_norm`: 1
198
- - `label_smoothing_factor`: 0.0
199
- - `bf16`: False
200
- - `fp16`: False
201
- - `bf16_full_eval`: False
202
- - `fp16_full_eval`: False
203
- - `tf32`: None
204
- - `gradient_checkpointing`: False
205
- - `gradient_checkpointing_kwargs`: None
206
- - `torch_compile`: False
207
- - `torch_compile_backend`: None
208
- - `torch_compile_mode`: None
209
- - `use_liger_kernel`: False
210
- - `liger_kernel_config`: None
211
- - `use_cache`: False
212
- - `neftune_noise_alpha`: None
213
- - `torch_empty_cache_steps`: None
214
- - `auto_find_batch_size`: False
215
- - `log_on_each_node`: True
216
- - `logging_nan_inf_filter`: True
217
- - `include_num_input_tokens_seen`: no
218
  - `log_level`: passive
219
  - `log_level_replica`: warning
220
- - `disable_tqdm`: False
221
- - `project`: huggingface
222
- - `trackio_space_id`: trackio
223
- - `eval_strategy`: no
224
- - `per_device_eval_batch_size`: 12
225
- - `prediction_loss_only`: True
226
- - `eval_on_start`: False
227
- - `eval_do_concat_batches`: True
228
- - `eval_use_gather_object`: False
229
- - `eval_accumulation_steps`: None
230
- - `include_for_metrics`: []
231
- - `batch_eval_metrics`: False
232
- - `save_only_model`: False
233
- - `save_on_each_node`: False
234
  - `enable_jit_checkpoint`: False
235
- - `push_to_hub`: False
236
- - `hub_private_repo`: None
237
- - `hub_model_id`: None
238
- - `hub_strategy`: every_save
239
- - `hub_always_push`: False
240
- - `hub_revision`: None
241
- - `load_best_model_at_end`: False
242
- - `ignore_data_skip`: False
243
  - `restore_callback_states_from_checkpoint`: False
244
- - `full_determinism`: False
245
  - `seed`: 42
246
  - `data_seed`: None
247
- - `use_cpu`: False
248
- - `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
249
- - `parallelism_config`: None
 
 
 
 
 
250
  - `dataloader_drop_last`: False
251
  - `dataloader_num_workers`: 0
252
- - `dataloader_pin_memory`: True
253
- - `dataloader_persistent_workers`: False
254
  - `dataloader_prefetch_factor`: None
 
255
  - `remove_unused_columns`: True
256
  - `label_names`: None
257
- - `train_sampling_strategy`: random
 
 
 
 
 
 
 
 
 
 
258
  - `length_column_name`: length
 
 
259
  - `ddp_find_unused_parameters`: None
260
  - `ddp_bucket_cap_mb`: None
261
  - `ddp_broadcast_buffers`: False
262
- - `ddp_backend`: None
263
- - `ddp_timeout`: 1800
264
- - `fsdp`: []
265
- - `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
266
- - `deepspeed`: None
267
- - `debug`: []
268
  - `skip_memory_metrics`: True
269
- - `do_predict`: False
270
  - `resume_from_checkpoint`: None
271
- - `warmup_ratio`: None
272
- - `local_rank`: -1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
273
  - `prompts`: None
274
  - `batch_sampler`: batch_sampler
275
  - `multi_dataset_batch_sampler`: round_robin
@@ -279,23 +987,95 @@ You can finetune this model on your own dataset.
279
  </details>
280
 
281
  ### Training Logs
282
- | Epoch | Step | Training Loss |
283
- |:------:|:----:|:-------------:|
284
- | 1.5152 | 500 | 0.0002 |
285
- | 3.0303 | 1000 | 0.0000 |
286
- | 4.5455 | 1500 | 0.0000 |
287
- | 0.7622 | 500 | 2.3552 |
288
- | 1.5244 | 1000 | 1.9437 |
289
- | 2.2866 | 1500 | 1.8059 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
290
 
291
 
292
  ### Framework Versions
293
- - Python: 3.12.7
294
  - Sentence Transformers: 5.2.3
295
- - Transformers: 5.3.0
296
- - PyTorch: 2.10.0
297
  - Accelerate: 1.13.0
298
- - Datasets: 4.2.0
299
  - Tokenizers: 0.22.2
300
 
301
  ## Citation
 
5
  - feature-extraction
6
  - dense
7
  - generated_from_trainer
8
+ - dataset_size:2369
9
  - loss:MultipleNegativesRankingLoss
10
+ - dataset_size:6288
11
  base_model: AI-Growth-Lab/PatentSBERTa
12
  widget:
13
+ - source_sentence: "COMPOSITIONS COMPRISING OR CONSISTING OF POLYDATIN FOR USE IN\
14
+ \ TREATMENT OF BONE DISEAS. Title: COMPOSITIONS COMPRISING OR CONSISTING OF POLYDATIN\
15
+ \ FOR USE IN TREATMENT OF BONE DISEAS\n\nAbstract: \n\nBusiness Description: The\
16
+ \ patent concerns the possible use of polydatin for the treatment of pathologies\
17
+ \ characterized by reduced bone mass. Polydatin is an oligostilbene that can be\
18
+ \ extracted in abundant quantities from the roots of the Polygonum Cuspidatum\
19
+ \ plant and, according to our studies about osteoblastic proliferation and differentiation,\
20
+ \ is able to increase osteogenic activity.\n\nTech Features: Skeletal pathologies\
21
+ \ are widespread in the world population, more than 200 million people are osteoporosis\
22
+ \ patients and have a high probability of fracture. Polydatin is a substance of\
23
+ \ natural origin, free of the side effects related to the use of synthetic drugs, currently\
24
+ \ administered for the treatment of bone loss diseases . Our experiments carried\
25
+ \ out in two-dimensional cultures (2D) and in cells grown on Scaffold (3D), have\
26
+ \ shown that polydatin is able to increase the activity of osteoblasts (ALP) and\
27
+ \ the deposition of Bone Matrix. The compositions comprising or consisting of\
28
+ \ polydatin may be administered orally (for example capsules or tablets, solutions,\
29
+ \ emulsions) or topically (oils, creams, ointments), or even through an appropriate\
30
+ \ enrichment of the active ingredient in functional foods\n\nApplications: Patients\
31
+ \ with traumatic-degenerative pathologies of the skeleton and oral cavity; Osteoporosis\
32
+ \ patients;\n\nAdvantages: Molecule of natural origin; Antioxidant activity on\
33
+ \ the whole organism; Possibility of topical use;"
34
  sentences:
35
+ - "Inhibitory compounds neurodegeneration and tumors. Title: Inhibitory compounds\
36
+ \ neurodegeneration and tumors\n\nAbstract: \n\nBusiness Description: The inhibition\
37
+ \ of the monoacylglycerol lipase enzyme (MAGL), naturally present in many brain\
38
+ \ cells and involved in physio-pathological processes, has a high therapeutic\
39
+ \ potential: neurodegenerative inflammation pathologies and tumors could be treated\
40
+ \ with new reversible inhibitory compounds, which would reduce the side effects\
41
+ \ of the irreversible inhibitors tested so far.\n\nTech Features: Monoacylglycerol\
42
+ \ lipase (MAGL) is a human enzyme of the endocannabinoid system involved in numerous\
43
+ \ physio-pathological processes (regulation of inflammation, anxiety, immune modulation,\
44
+ \ motor coordination ...), yet its overexpression/upregulation can cause neuroinflammatory\
45
+ \ diseases and tumors. The inhibition of MAGL for therapeutic purposes has been\
46
+ \ studied so far with irreversible inhibitors, which however nullify the enzyme\
47
+ \ activity, leading to a progressive loss of the therapeutic effect and to addiction\
48
+ \ phenomena. On the contrary, the new-patented compounds based on a strong non-covalent\
49
+ \ reversible mechanism of action avoid the side effects mentioned. Effective in\
50
+ \ laboratory on various tumor cell lines (e.g. colorectal, breast and ovarian\
51
+ \ cancer), they could also treat other MAGL-mediated pathologies (neuroinflammation/degeneration,\
52
+ \ pain, amyotrophic multiple/lateral sclerosis, Alzheimer's disease, Parkinson's\
53
+ \ disease). Link to scientific publication and information on Ca' Foscari website\n\
54
+ \nApplications: Innovative and less harmful pharmaceutical compositions for the\
55
+ \ treatment of serious neurodegenerative pathological conditions; Innovative and\
56
+ \ less harmful pharmaceutical compositions for cancer treatment.\n\nAdvantages:\
57
+ \ Temporary nature and reversibility; Drastic reduction of side effects; High\
58
+ \ efficacy tested on tumor cell lines; Exploitable to treat numerous neurodegenerative\
59
+ \ diseases; One of the few non-covalent reversible MAGL inhibitors with high efficacy."
60
+ - 'New computational method for the prognosis of amyotrophic lateral sclerosis.
61
+ Title: New computational method for the prognosis of amyotrophic lateral sclerosis
62
+
63
+
64
+ Abstract: A method is described for determining a disease progression and survival
65
+ prognosis, at a succession of prediction times, for patients suffering from amyotrophic
66
+ lateral sclerosis (ALS). The method comprises a step of defining a set of variables
67
+ associated with the onset and progression of amyotrophic lateral sclerosis, comprising
68
+ a first group of variables associated with the onset of amyotrophic lateral sclerosis
69
+ (comprising at least the variables “patient sex”, “disease onset age”, “disease
70
+ onset site”), a second group of dynamic time variables (comprising at least the
71
+ variable “time elapsed since disease onset”), a third group of dynamic functional
72
+ variables (comprising at least one of the variables breathing, swallowing, communicating,
73
+ walking/self-care or at least one variable of a functional progression and/or
74
+ severity scale of amyotrophic lateral sclerosis), and further at least one variable
75
+ associated with survival. The method further provides for encoding by means of
76
+ a Dynamic Bayesian Network, using at least one trained algorithm, a plurality
77
+ of probabilistic conditional dependence relationships, in which each relationship
78
+ is a probabilistic conditional dependence relationship between two of the aforesaid
79
+ variables. The aforesaid prediction times are defined so that each prediction
80
+ time belongs to a respective time interval in which the conditional dependence
81
+ relationships between the variables are stationary. The method further involves
82
+ describing the Dynamic Bayesian Network, using at least one trained algorithm,
83
+ by means of a corresponding graph, comprising said variables as nodes and comprising
84
+ topological connections oriented between nodes corresponding to variables among
85
+ which a probabilistic conditional dependence is identified. In the graph, given
86
+ a node, the connections entering it show a conditional probability of the value
87
+ assumed by the variable associated with such node, in a given prediction time,
88
+ depending on the values assumed, in a prior prediction time, from the variables
89
+ associated with the nodes from which such connections originate. The method further
90
+ comprises the steps of entering, for each of the defined variables, data acquired
91
+ at a given acquisition time relating to the situation of a specific patient; and
92
+ calculating, by electronic processing and/or calculating means, on the basis of
93
+ the Dynamic Bayesian Network and the graph, and starting from the aforesaid acquired
94
+ data, the values of each of the defined variables, at one or more prediction times
95
+ following the acquisition time. Finally, the method involves obtaining, in a given
96
+ prediction time, disease progression prognosis results on the basis of the values
97
+ of one or more of the variables of the third group calculated in such prediction
98
+ time; and the survival prognosis results on the basis of the value of at least
99
+ one variable associated with survival, calculated at such prediction time.
100
+
101
+
102
+ Business Description: The invention relates to a new probabilistic model based
103
+ on dynamic Bayesian networks: by introducing specific clinical data of patients
104
+ suffering from amyotrophic lateral sclerosis (ALS), the model object of the invention
105
+ allows to predict the course of the disease, the loss of autonomy in specific
106
+ functional domains and the survival, and to identify risk factors of the disease.
107
+
108
+
109
+ Tech Features: Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease,
110
+ characterized by progressive muscle paralysis caused by motor neuron degeneration.
111
+ The onset and progression of ALS are very heterogeneous, making the development
112
+ of personalised approaches extremely difficult. The invention consists of a probabilistic
113
+ predictor of ALS progression, based on a dynamic Bayesian network built using
114
+ a database of over 4000 patients treated at specialised centres in Italy and Israel.
115
+ By intoducing the clinical information of a new ALS subject, the invention allows
116
+ to simulate the clinical course and to probabilistically predict the risk of impairment
117
+ of the functional domains typical of ALS and the survival, representing an important
118
+ tool to support clinical decision-making and the drafting of the treatment plan.
119
+
120
+
121
+ Applications: Online platform (SaaS)/software for predicting ALS progression;
122
+ Support to clinical decision and drafting of the treatment plan; Institutional
123
+ stakeholder decision support; In silico generation of patient populations with
124
+ specific characteristics.
125
+
126
+
127
+ Advantages: Prediction of ALS progression; Personalized medicine approach; Clinical
128
+ decision support; Identification of new prognostic markers; Possible alternative
129
+ to placebo cohorts in clinical trials.'
130
+ - "COMPOSITIONS COMPRISING OR CONSISTING OF POLYDATIN FOR USE IN TREATMENT OF BONE\
131
+ \ DISEAS. Title: COMPOSITIONS COMPRISING OR CONSISTING OF POLYDATIN FOR USE IN\
132
+ \ TREATMENT OF BONE DISEAS\n\nAbstract: \n\nBusiness Description: The patent concerns\
133
+ \ the possible use of polydatin for the treatment of pathologies characterized\
134
+ \ by reduced bone mass. Polydatin is an oligostilbene that can be extracted in\
135
+ \ abundant quantities from the roots of the Polygonum Cuspidatum plant and, according\
136
+ \ to our studies about osteoblastic proliferation and differentiation, is able\
137
+ \ to increase osteogenic activity.\n\nTech Features: Skeletal pathologies are\
138
+ \ widespread in the world population, more than 200 million people are osteoporosis\
139
+ \ patients and have a high probability of fracture. Polydatin is a substance of\
140
+ \ natural origin, free of the side effects related to the use of synthetic drugs, currently\
141
+ \ administered for the treatment of bone loss diseases . Our experiments carried\
142
+ \ out in two-dimensional cultures (2D) and in cells grown on Scaffold (3D), have\
143
+ \ shown that polydatin is able to increase the activity of osteoblasts (ALP) and\
144
+ \ the deposition of Bone Matrix. The compositions comprising or consisting of\
145
+ \ polydatin may be administered orally (for example capsules or tablets, solutions,\
146
+ \ emulsions) or topically (oils, creams, ointments), or even through an appropriate\
147
+ \ enrichment of the active ingredient in functional foods\n\nApplications: Patients\
148
+ \ with traumatic-degenerative pathologies of the skeleton and oral cavity; Osteoporosis\
149
+ \ patients;\n\nAdvantages: Molecule of natural origin; Antioxidant activity on\
150
+ \ the whole organism; Possibility of topical use;"
151
+ - source_sentence: "OFFLOADNN. Title: OFFLOADNN\n\nAbstract: \n\nBusiness Description:\
152
+ \ Resource availability at the edge is generally limited, with particular regard\
153
+ \ to memory consumption of DNNs, rarely considered as a limiting factor to the\
154
+ \ execution of tasks at the edge. Furthermore, the structure of the DNNs required\
155
+ \ for different CV inference tasks and the correlation among them have often been\
156
+ \ overlooked. Accounting for all this, the invention decreases the resource consumption\
157
+ \ of DNNs dedicated to run CV tasks, which in turn allow to run more tasks or\
158
+ \ to decrease the energy consumption with the same number of running tasks.\n\n\
159
+ Tech Features: The invention is OffloaDNN, a framework to support the scalable\
160
+ \ offloading of computer vision (CV) tasks to the edge. We formulated the problem\
161
+ \ of DNN for scalable Offloading of Tasks (DOT) to find the optimal configuration\
162
+ \ of the DNN, the tasks that can be admitted and the allocation of compute and\
163
+ \ network resources to assign to the tasks. DOT minimizes the task rejection rate\
164
+ \ and resource consumption, while meeting the accuracy and latency requirements\
165
+ \ of admitted tasks. As the DOT is NP-hard, and therefore complex to solve to\
166
+ \ optimality, OffloaDNN solves the DOT efficiently using a weighted tree-based\
167
+ \ graph modeling of the feasible solutions. The invention allows to decrease the\
168
+ \ resources needed at the edge, also allowing for multiple task execution when\
169
+ \ these require the same DNN components. The lowering of resources needed implies\
170
+ \ also a lower energy consumption with operational savings.\n\nApplications: Computer\
171
+ \ vision tasks using edge computing.\n\nAdvantages: Lower resources needed at\
172
+ \ the edge; Lower energy consumption; Lower operative costs."
173
  sentences:
174
+ - "OFFLOADNN. Title: OFFLOADNN\n\nAbstract: \n\nBusiness Description: Resource availability\
175
+ \ at the edge is generally limited, with particular regard to memory consumption\
176
+ \ of DNNs, rarely considered as a limiting factor to the execution of tasks at\
177
+ \ the edge. Furthermore, the structure of the DNNs required for different CV inference\
178
+ \ tasks and the correlation among them have often been overlooked. Accounting\
179
+ \ for all this, the invention decreases the resource consumption of DNNs dedicated\
180
+ \ to run CV tasks, which in turn allow to run more tasks or to decrease the energy\
181
+ \ consumption with the same number of running tasks.\n\nTech Features: The invention\
182
+ \ is OffloaDNN, a framework to support the scalable offloading of computer vision\
183
+ \ (CV) tasks to the edge. We formulated the problem of DNN for scalable Offloading\
184
+ \ of Tasks (DOT) to find the optimal configuration of the DNN, the tasks that\
185
+ \ can be admitted and the allocation of compute and network resources to assign\
186
+ \ to the tasks. DOT minimizes the task rejection rate and resource consumption,\
187
+ \ while meeting the accuracy and latency requirements of admitted tasks. As the\
188
+ \ DOT is NP-hard, and therefore complex to solve to optimality, OffloaDNN solves\
189
+ \ the DOT efficiently using a weighted tree-based graph modeling of the feasible\
190
+ \ solutions. The invention allows to decrease the resources needed at the edge,\
191
+ \ also allowing for multiple task execution when these require the same DNN components.\
192
+ \ The lowering of resources needed implies also a lower energy consumption with\
193
+ \ operational savings.\n\nApplications: Computer vision tasks using edge computing.\n\
194
+ \nAdvantages: Lower resources needed at the edge; Lower energy consumption; Lower\
195
+ \ operative costs."
196
+ - 'Polymeric endodontic instruments. Title: Polymeric endodontic instruments
197
+
198
+
199
+ Abstract: The present invention relates to an endodontic tool for shaping a dental
200
+ root canal, wherein said tool comprises a handle (13), a tang (11), and a stem
201
+ (10) made of polymeric material and comprising a tip (12) and at least one helically
202
+ developing cutting edge (100), and wherein said stem (10) develops along a longitudinal
203
+ axis (A-A). The stem (10) has a substantially conical or tapering twisted shape,
204
+ in particular said stem (10) having a solid cross-section with a diameter increasing
205
+ from the tip (12) to the tang (11).
206
+
207
+
208
+ Business Description: Endodontic files for shaping canal roots are made of Ni-Ti
209
+ alloy or stainless steel. Production, based on precision micromechanics processes,
210
+ involves high costs and time. Production limitations affect the geometries, especially
211
+ near the tip, reducing cutting efficiency. Another critical issue is the potential
212
+ failure of the instrument during treatment, which can trap it in the canal and
213
+ require invasive procedures for the patient.
214
+
215
+
216
+ Tech Features: The analysis of materials databases has identified families of
217
+ biocompatible polymers, potentially reinforced, capable of sustaining the applied
218
+ loads even better than the traditional metallic materials used for endodontic
219
+ files. The selected polymers offer advantages in terms of fatigue resistance and
220
+ reduced costs. However, they present critical issues related to hardness and torsion
221
+ resistance, which are fundamental characteristics for the shaping of the canal
222
+ root. To address these issues, the use of biocompatible surface coatings with
223
+ high hardness and low friction coefficient is proposed, applicable through technologies
224
+ such as chemical vapor deposition. These coatings can significantly improve the
225
+ surface resistance of polymers, making them suitable for use in endodontic files.
226
+
227
+
228
+ Applications: Biocompatible polymers have been identified that withstand critical
229
+ stresses leading to the failure of metallic tools during canal shaping; The identified
230
+ polymers, with lower density and cost compared to metals, allow the use of inexpensive
231
+ mass production technologies.
232
+
233
+
234
+ Advantages: Drastic reduction in production costs; Reduction or elimination of
235
+ the risk of instrument failure in the canal; Creation of innovative geometries
236
+ through 3D printing or injection molding for large-scale production.'
237
+ - 'Unidirectional mechanical device. Title: Unidirectional mechanical device
238
+
239
+
240
+ Abstract: Unidirectional mechanical device comprising at least one array (10;
241
+ 20) of three or more cells (100; 200). Each cell (100; 200) of the array comprises:
242
+ a floating mass (101; 201); at least one first elastic element (102; 202) connecting
243
+ the floating mass (101; 201) to a reference external structure for the unidirectional
244
+ mechanical device, wherein the at least one first elastic element (102; 202) comprises
245
+ at least one first folded beam; at least one second elastic element (103) connecting
246
+ the floating mass (101) to a second floating mass of a preceding cell of the at
247
+ least one array (10), wherein the at least one second elastic element (103) comprises
248
+ at least one second folded beam; at least one third elastic element (104) connecting
249
+ the floating mass (101) to a third floating mass of a succeeding cell of the at
250
+ least one array (10), wherein the at least one third elastic element (104) comprises
251
+ at least one third folded beam; an electric system configured to control a potential
252
+ difference acting on the floating mass (101; 201), determining an overall electrostatic
253
+ stiffness of the at least one array (10; 20).
254
+
255
+
256
+ Business Description: The invention consists of a mono-directional mechanical
257
+ filter at a micrometric scale, which can be integrated into MEMS systems. In certain
258
+ frequency bands, the device allows the propagation of mechanical waves in only
259
+ one direction, blocking those travelling in the opposite direction. We obtain
260
+ this filter by means of a periodic structure with time-modulated stiffness, here
261
+ achieved by regulating suitable voltages.
262
+
263
+
264
+ Tech Features: In mechanics, the principle of reciprocity dictates that a system
265
+ excited at point A and observed at point B behaves in the same way if excited
266
+ at B and observed at A. One of the ways to create a non-reciprocal system is to
267
+ break temporal invariance. A substantial number of scientific works in recent
268
+ years have dealt with the theoretical and experimental description of these systems,
269
+ created through periodic structures of time-varying stiffness. However, these
270
+ works remain confined to the level of speculation and demonstrative prototypes
271
+ to the macro-scale, with complicated and difficult to implement solutions. The
272
+ present invention realizes the above principles in an easy-to-produce MEMS device
273
+ by using an electrostatic principle for the modulation of the stiffness which,
274
+ since it is not present at the macro-scale but only at the micro-scale, had not
275
+ been considered with this aim up to now. Essentially, the system is made up of
276
+ the repetition of some elementary cells, made up of springs and masses that house
277
+ an electrode with parallel faces for modulation. We speculate that this new type
278
+ of system can be used both as a new sensor and/or as a component (e.g. filter)
279
+ in more complex systems.
280
+
281
+
282
+ Applications: MEMS directional sensor; Directional filter for devices that use
283
+ elastic waves as the main means of transduction.
284
+
285
+
286
+ Advantages: Uses mechanical waves; Mechanical device with non-symmetrical response
287
+ (equivalent of an electric diode); Lower frequencies as compared to electromagnetic
288
+ waves; Can be manufactured using traditional MEMS technologies.'
289
+ - source_sentence: "GEL FORMULATION FOR THE ORAL ADMINISTRATION OF DRUGS IN PARTICULAR\
290
+ \ TO DYSPHAGIC PATIENTS. Title: GEL FORMULATION FOR THE ORAL ADMINISTRATION OF\
291
+ \ DRUGS IN PARTICULAR TO DYSPHAGIC PATIENTS\n\nAbstract: \n\nBusiness Description:\
292
+ \ The present invention relates to a gel formulation containing water-soluble\
293
+ \ active ingredients mainly available on the pharmaceutical market as oral solid\
294
+ \ dosage forms such as immediate-release tablets or capsules or powders or granules\
295
+ \ for oral use. The formulation object of the present invention satisfies the\
296
+ \ particular needs of patients who show anomalies in the swallowing process. The\
297
+ \ pathological condition related to swallowing problems is called dysphagia.\n\
298
+ \nTech Features: The present invention provides a method for reformulating water-soluble\
299
+ \ drugs, when commercially available for oral administration only in solid form\
300
+ \ with immediate release, mostly tablets and / or capsules without neglecting\
301
+ \ the existence of granules and powders for oral use. Drugs are reformulated as\
302
+ \ semi-solid preparations suitable for oral administration in patients with swallowing.\
303
+ \ This method allows to obtain an unchanged release of the active ingredient,\
304
+ \ compared to that of the drug formulated in the original dosage form. The present\
305
+ \ patent therefore describes a semi-solid formulation with rheological characteristics\
306
+ \ comparable to commercially gelled drinks, used for hydration and the administration\
307
+ \ of therapy to patients with swallowing disorders.\n\nApplications: Avoid any\
308
+ \ handling operations by eliminating the dangers associated with loss of active\
309
+ \ substance; Avoid contamination and dosage errors with greater safety in the\
310
+ \ administration of drugs; The active substance is formulated in a dosage form\
311
+ \ with a consistency suitable for swallowing in the dysphagic patients without\
312
+ \ its release kinetics being altered.\n\nAdvantages: Potential applicability in\
313
+ \ hospital pharmacies to standardize, optimize and automate the galenic preparations\
314
+ \ of drugs available only as tablets; Large-scale marketing can improve the therapeutic\
315
+ \ possibilities for dysphagia patients; Industrial production may cover a market\
316
+ \ share, which is currently entrusted to galenic production in a pharmacy."
317
  sentences:
318
+ - "Supplement for psychiatric disorders. Title: Supplement for psychiatric disorders\n\
319
+ \nAbstract: \n\nBusiness Description: The present invention relates to an ester\
320
+ \ of a phospholipid with conjugated linoleic acid (CLA) as a nutritional supplement\
321
+ \ in treatment of psychiatric disorders with neuroinflammatory and neurodegenerative\
322
+ \ basis, such as depression, bipolar disorder and schizophrenia.\n\nTech Features:\
323
+ \ The search for alternative therapies in place of non-steroidal anti-inflammatory\
324
+ \ drugs for neuroinflammation is crucial since the risk of side effects derived\
325
+ \ from their use has been demonstrated especially during pregnancy with serious\
326
+ \ outcomes such as the development of sexual organs in the fetus as well as subsequent\
327
+ \ psychiatric trajectory in the offspring life in both sexes. Therefore, an alternative\
328
+ \ therapeutic approach would be resolutive to pharmacological treatment. We have\
329
+ \ shown in in-vivo models that the dietary intake of 0.5% CLA (i) increases omega-3\
330
+ \ DHA fatty acid levels by approx. 10 times in both the maternal and fetal liver,\
331
+ \ and approx. 5 times in the fetal brain; (ii) increases gene expression of PPAR\
332
+ \ alpha and its PEA and OEA ligands. These molecules are known to exert a potent\
333
+ \ anti-neuroinflammatory activity and thereby a protective effect against those\
334
+ \ psychiatric disorders with neuroinflammatory basis. Furthermore, as ester of\
335
+ \ a phospholipid the bioavailability of CLA is increased.\nTRL= 3 - 4 ( in-vivo\
336
+ \ )\n\nApplications: Psychiatric disorders with neuroinflammatory basis such as\
337
+ \ depression, bipolar disorder and schizophrenia.\n\nAdvantages: Absence of side\
338
+ \ effects; Suitable also during critical windows of vulnerability such as pregnancy\
339
+ \ and infancy; CLA can be formulated as a functional food, facilitating its compliance."
340
+ - "Safepack: Safety Backpack for riders. Title: Safepack: Safety Backpack for riders\n\
341
+ \nAbstract: \n\nBusiness Description: Safepack is the first smart backpack designed\
342
+ \ for Gig Economy delivery service riders. It is fully integrated into the work\
343
+ \ dynamics and improves user safety by reducing the possibility of distraction\
344
+ \ and enhancing communication with other road users.\n\nTech Features: Safepack\
345
+ \ consists of: 1) Tactile navigation system, inserted in the belt, consisting\
346
+ \ of piezoelectric actuators that vibrate in the direction of the path to follow\
347
+ \ and an inflation system that ensures the fitting and correct perception of the\
348
+ \ vibration. 2) Communication system based on voice interface (Speaker and microphone):\
349
+ \ the rider can receive information from the platform, such as the delivery of\
350
+ \ a new order (collection / delivery point, distance, compensation), can accept\
351
+ \ or reject it and communicate delays or emergencies. 3) Automatic lighting system,\
352
+ \ aimed at improving communication with other road users. The light is generated\
353
+ \ by side-emitting optical fibers connected to a yellow LED for the visibility\
354
+ \ lights, and red LEDs for the directional and brake lights. Finally, Safepack\
355
+ \ communicates via bluetooth with the platform app and with the navigation app.\
356
+ \ The volume of the voice interface and the intensity of the vibration can be\
357
+ \ adjusted thanks to a pressure sensor placed in the right shoulder strap.\n\n\
358
+ Applications: Riders working in Gig economy delivery services.\n\nAdvantages:\
359
+ \ A smart product perfectly integrated into the working dynamics of the delivery\
360
+ \ platforms; Offers a support tool that allows a rider not to have to constantly\
361
+ \ look at a smartphone to receive navigation instructions and information from\
362
+ \ the platform app; Improves rider safety by reducing distracting factors, making\
363
+ \ him visible and communicating his movements to other road users."
364
+ - "GEL FORMULATION FOR THE ORAL ADMINISTRATION OF DRUGS IN PARTICULAR TO DYSPHAGIC\
365
+ \ PATIENTS. Title: GEL FORMULATION FOR THE ORAL ADMINISTRATION OF DRUGS IN PARTICULAR\
366
+ \ TO DYSPHAGIC PATIENTS\n\nAbstract: \n\nBusiness Description: The present invention\
367
+ \ relates to a gel formulation containing water-soluble active ingredients mainly\
368
+ \ available on the pharmaceutical market as oral solid dosage forms such as immediate-release\
369
+ \ tablets or capsules or powders or granules for oral use. The formulation object\
370
+ \ of the present invention satisfies the particular needs of patients who show\
371
+ \ anomalies in the swallowing process. The pathological condition related to swallowing\
372
+ \ problems is called dysphagia.\n\nTech Features: The present invention provides\
373
+ \ a method for reformulating water-soluble drugs, when commercially available\
374
+ \ for oral administration only in solid form with immediate release, mostly tablets\
375
+ \ and / or capsules without neglecting the existence of granules and powders for\
376
+ \ oral use. Drugs are reformulated as semi-solid preparations suitable for oral\
377
+ \ administration in patients with swallowing. This method allows to obtain an\
378
+ \ unchanged release of the active ingredient, compared to that of the drug formulated\
379
+ \ in the original dosage form. The present patent therefore describes a semi-solid\
380
+ \ formulation with rheological characteristics comparable to commercially gelled\
381
+ \ drinks, used for hydration and the administration of therapy to patients with\
382
+ \ swallowing disorders.\n\nApplications: Avoid any handling operations by eliminating\
383
+ \ the dangers associated with loss of active substance; Avoid contamination and\
384
+ \ dosage errors with greater safety in the administration of drugs; The active\
385
+ \ substance is formulated in a dosage form with a consistency suitable for swallowing\
386
+ \ in the dysphagic patients without its release kinetics being altered.\n\nAdvantages:\
387
+ \ Potential applicability in hospital pharmacies to standardize, optimize and\
388
+ \ automate the galenic preparations of drugs available only as tablets; Large-scale\
389
+ \ marketing can improve the therapeutic possibilities for dysphagia patients;\
390
+ \ Industrial production may cover a market share, which is currently entrusted\
391
+ \ to galenic production in a pharmacy."
392
+ - source_sentence: 'TARGETING SMALL RNAS AS A THERAPY FOR ALS. Title: TARGETING SMALL
393
+ RNAS AS A THERAPY FOR ALS
394
+
395
+
396
+ Abstract: The present invention relates to an inhibitor of miR-129, relative compounds
397
+ and pharmaceutical compositions for use in the treatment and/or prevention of
398
+ amyotrophic lateral sclerosis and Alzheimer''s disease. The invention also relates
399
+ to a method for the diagnosis and/or prognosis of Alzheimer''s disease in a subject
400
+ or to identify a subject at risk to develop amyotrophic lateral sclerosis or Alzheimer''s
401
+ disease and to a method for the measuring the efficacy of a therapy for amyotrophic
402
+ lateral sclerosis or for Alzheimer''s disease and relative kits.
403
+
404
+
405
+ Business Description: We propose to develop a drug product to treat all forms
406
+ of ALS by targeting an important pathogenic pathway, miR-129-1 that is up-regulated
407
+ in both familial and sporadic ALS patients, in human and SOD1 G93A mice (A-B).
408
+ Antisense technology is entering a phase of clinical successes and we believe
409
+ that our strategy can provide clinically meaningful benefit to patients that have
410
+ no therapeutic hope. To date, current approved treatments for ALS, which at best
411
+ guarantee a span life extension of 3 months, are riluzole (Li et al., Plos One
412
+ 2013) and edaravone (Ito et al, Exp Neurol, 2008) .
413
+
414
+
415
+ Tech Features: Our proposed miR-129-1 PMO therapeutic strategy outperformed when
416
+ compared with these ALS approved molecules. Our patented a miR-129-1 modulation
417
+ strategy with a Morpholino Antisense Oligonucleotide (ASO-PMO) and demonstrated
418
+ in vivo that it recovers the expression of its target protein HuD, allowing for
419
+ a 12.5% ​​increase in survival and an improvement neuromuscular performance of
420
+ SOD1 G93A mice, a genetic model of ALS, treated in the pre-symptomatic phase. To
421
+ improve miR-129-1 PMO therapeutic efficacy, better delivery to target tissue is
422
+ needed: based on our experimental data from another study on motor neuron disease,
423
+ an arginine-rich peptide conjugated with PMO (CPP-PMO) shows an increased efficacy
424
+ in vivo . In addition, the mouse model TDP43 can be used to demonstrate a more
425
+ general efficacy of the product, before proceeding with toxicological and pharmacokinetic
426
+ evaluation in the preclinical setting. In addition, the mouse model TDP43 can
427
+ be used to demonstrate a more general efficacy of the product, before proceeding
428
+ with toxicological and pharmacokinetic evaluation in the preclinical setting.
429
+
430
+
431
+ Applications: Familial amyotrophic lateral sclerosis, Sporadic amyotrophic lateral
432
+ sclerosis, Alzheimer Disease, Epilepsy, Fragile X syndrome, Tubero sclerosis complex
433
+ and Cornea pathologies.
434
+
435
+
436
+ Advantages: Targeting miRNA allows modulation of a large number of target genes;
437
+ PMO works as a steric-blocker, a more effective strategy when the target is a
438
+ miRNA; PMO is superior to small molecules for improved target affinity; PMO fulfills
439
+ important criteria of solubility, stability, binding affinity, in vitro and in
440
+ vivo efficacy and in vivo toxicity if compared to other chemistries; PMO has low
441
+ risk in terms of toxicity, off target effects and immune response stimulation.'
442
  sentences:
443
+ - 'Antimicrobial dental adhesive graphene-based. Title: Antimicrobial dental adhesive
444
+ graphene-based
445
+
446
+
447
+ Abstract: Disclosed is a dental adhesive including a polymeric adhesive and a
448
+ nanofiller dispersed in the polymeric adhesive, the nanofiller being constituted
449
+ by graphene nanostructures which are properly dispersed inside the polymer adhesive
450
+ and over the surface of the adhesive layer without formation of agglomerates,
451
+ so that the dental adhesive exhibits significant antimicrobial and antibiofilm
452
+ properties against pathogens of the oral cavity.
453
+
454
+
455
+ Business Description: The patent is based on a dental adhesive with antimicrobial
456
+ and anti-biofilm properties taking advantage of the use of graphene-based micro
457
+ and nanometer fillers. The nanomaterials introduced into an adhesive are uniformly
458
+ dispersed, without agglomerations inside the polymer and partially exposed over
459
+ the adhesive surface. These significantly inhibit, by mechanical action, adhesion
460
+ and bacterial growth.
461
+
462
+
463
+ Tech Features: The adhesive object of the patent consists of a dental adhesive
464
+ composed of BisGMA, filled with graphene-based nanoplachets (GNP), produced at
465
+ low cost by thermal expansion, or its derivatives incorporated in the polymer
466
+ matrix. Adhesives have strong antimicrobial properties against the bacteria that
467
+ typically populate the oral cavity, such as Streptococcus mutants. Current commercial
468
+ adhesives use nanometric fillers, in particular barium glass, zirconium oxide,
469
+ aluminum oxide, or silicon dioxide, for the sole purpose of increasing the micromechanical
470
+ adhesion between the adhesive and dentin or enamel, and therefore typically do
471
+ not have specific properties. antimicrobials. The antimicrobial dental adhesives
472
+ used have limitations, as organic fillers tend to worsen the mechanical characteristics
473
+ of dental materials. In fact, these fillers are subject to shrinkage after the
474
+ cure and polymerization phase, just like the matrix and therefore limit the mechanical
475
+ strength of the composite.
476
+
477
+
478
+ Applications: Dental adhesive with antimicrobial properties.
479
+
480
+
481
+ Advantages: The particular production process of the new adhesive possess antimicrobial
482
+ properties; The production of this adhesive does not require structural changes
483
+ to companies producing dental adhesives; GNP possess the properties of the basal
484
+ plane typical of 2D graphene, at a lower cost and without its technological problems.
485
+ Moreover they can be used on a large scale.'
486
+ - "OXYSTEROL FOR THE TREATMENT AND PREVENTION OF CORONAVIRUS DISEASES. Title: OXYSTEROL\
487
+ \ FOR THE TREATMENT AND PREVENTION OF CORONAVIRUS DISEASES\n\nAbstract: \n\nBusiness\
488
+ \ Description: The invention aims to provide antiviral compounds for the prevention\
489
+ \ and treatment of infections caused by a coronavirus, particularly by SARS-CoV-2.\
490
+ \ The invention shows that 27-hydroxycholesterol can activate the innate immune\
491
+ \ response and induce the response of cells and tissues. Formulated in a pharmaceutical\
492
+ \ composition, it significantly inhibits the infectivity of SARS-CoV-2 (100% at\
493
+ \ the highest doses).\n\nTech Features: An oxysterol called 27-hydroxycholesterol\
494
+ \ (27OHC), a derivative of cholesterol obtained by oxidation, endogenous, that\
495
+ \ is, produced naturally by the human body, has been identified as an antiviral\
496
+ \ compound. In fact,  in vitro tests, 27OHC was administered after inoculating\
497
+ \ susceptible cells with SARS-CoV-2, simulating a viral infection treatment protocol.\
498
+ \ 27OHC was able to inhibit the replication of SARS-CoV-2 and other human coronavirus\
499
+ \ that causes the common cold (HCoV-OC43), without cytotoxicity, at concentrations\
500
+ \ in the micromolar range.\n\nThis study has been published on the journal Redox\
501
+ \ Biology and is available ( https://doi.org/10.1016/j.redox.2020.101682 ).\n\n\
502
+ Applications: Pharmaceutical sector; Respiratory diseases; Chronic inflammatory\
503
+ \ gastrointestinal diseases.\n\nAdvantages: New treatment for SARS-CoV-2; Treatment\
504
+ \ active against other Coronaviruses; Physiological drug naturally present in\
505
+ \ the human organism."
506
+ - 'TARGETING SMALL RNAS AS A THERAPY FOR ALS. Title: TARGETING SMALL RNAS AS A THERAPY
507
+ FOR ALS
508
+
509
+
510
+ Abstract: The present invention relates to an inhibitor of miR-129, relative compounds
511
+ and pharmaceutical compositions for use in the treatment and/or prevention of
512
+ amyotrophic lateral sclerosis and Alzheimer''s disease. The invention also relates
513
+ to a method for the diagnosis and/or prognosis of Alzheimer''s disease in a subject
514
+ or to identify a subject at risk to develop amyotrophic lateral sclerosis or Alzheimer''s
515
+ disease and to a method for the measuring the efficacy of a therapy for amyotrophic
516
+ lateral sclerosis or for Alzheimer''s disease and relative kits.
517
+
518
+
519
+ Business Description: We propose to develop a drug product to treat all forms
520
+ of ALS by targeting an important pathogenic pathway, miR-129-1 that is up-regulated
521
+ in both familial and sporadic ALS patients, in human and SOD1 G93A mice (A-B).
522
+ Antisense technology is entering a phase of clinical successes and we believe
523
+ that our strategy can provide clinically meaningful benefit to patients that have
524
+ no therapeutic hope. To date, current approved treatments for ALS, which at best
525
+ guarantee a span life extension of 3 months, are riluzole (Li et al., Plos One
526
+ 2013) and edaravone (Ito et al, Exp Neurol, 2008) .
527
+
528
+
529
+ Tech Features: Our proposed miR-129-1 PMO therapeutic strategy outperformed when
530
+ compared with these ALS approved molecules. Our patented a miR-129-1 modulation
531
+ strategy with a Morpholino Antisense Oligonucleotide (ASO-PMO) and demonstrated
532
+ in vivo that it recovers the expression of its target protein HuD, allowing for
533
+ a 12.5% ​​increase in survival and an improvement neuromuscular performance of
534
+ SOD1 G93A mice, a genetic model of ALS, treated in the pre-symptomatic phase. To
535
+ improve miR-129-1 PMO therapeutic efficacy, better delivery to target tissue is
536
+ needed: based on our experimental data from another study on motor neuron disease,
537
+ an arginine-rich peptide conjugated with PMO (CPP-PMO) shows an increased efficacy
538
+ in vivo . In addition, the mouse model TDP43 can be used to demonstrate a more
539
+ general efficacy of the product, before proceeding with toxicological and pharmacokinetic
540
+ evaluation in the preclinical setting. In addition, the mouse model TDP43 can
541
+ be used to demonstrate a more general efficacy of the product, before proceeding
542
+ with toxicological and pharmacokinetic evaluation in the preclinical setting.
543
+
544
+
545
+ Applications: Familial amyotrophic lateral sclerosis, Sporadic amyotrophic lateral
546
+ sclerosis, Alzheimer Disease, Epilepsy, Fragile X syndrome, Tubero sclerosis complex
547
+ and Cornea pathologies.
548
+
549
+
550
+ Advantages: Targeting miRNA allows modulation of a large number of target genes;
551
+ PMO works as a steric-blocker, a more effective strategy when the target is a
552
+ miRNA; PMO is superior to small molecules for improved target affinity; PMO fulfills
553
+ important criteria of solubility, stability, binding affinity, in vitro and in
554
+ vivo efficacy and in vivo toxicity if compared to other chemistries; PMO has low
555
+ risk in terms of toxicity, off target effects and immune response stimulation.'
556
+ - source_sentence: "High reduction and backdrivability ACTUATion SYSTEM with minimal\
557
+ \ lateral encumbrance. Title: High reduction and backdrivability ACTUATion SYSTEM\
558
+ \ with minimal lateral encumbrance\n\nAbstract: \n\nBusiness Description: Typical\
559
+ \ robotic actuation units have a motor and its transmission system aligned with\
560
+ \ the axis of the joint they actuate, resulting in large lateral encumbrance of\
561
+ \ the whole system.\n\nThis solution aims at solving this problem and allows to\
562
+ \ locate the last transmission stage at an arbitrary distance from the axis of\
563
+ \ the screw.\n\nTech Features: This invention consists in a class of actuation\
564
+ \ systems which is specifically designed to minimize the lateral encumbrance of\
565
+ \ the exoskeletal system to maximize its practical usability. Its core components\
566
+ \ are a motor coupled to a leadscrew or ballscrew system and a further (arbitrary)\
567
+ \ transmission system to connect the nut of the screw based transmission to the\
568
+ \ output wheel.\n\nFurthermore, this solution allows to displace the location\
569
+ \ of the screw with respect to the final transmission stage, allowing the adaptation\
570
+ \ of the location of the different stages of the system without influencing the\
571
+ \ overall behavior.The designed anti-blockage system guarantees proper functioning\
572
+ \ of each mechanical component and high back-drivability of the overall system\
573
+ \ even for high overall gearing.\n\nApplications: Robotics.\n\nAdvantages: Minimal\
574
+ \ lateral encumbrance; High gearing; Highly Back-drivable; Torque estimation through\
575
+ \ current measurement; Capability of dislocating the screw from the point of application\
576
+ \ of force."
577
  sentences:
578
+ - "High reduction and backdrivability ACTUATion SYSTEM with minimal lateral encumbrance.\
579
+ \ Title: High reduction and backdrivability ACTUATion SYSTEM with minimal lateral\
580
+ \ encumbrance\n\nAbstract: \n\nBusiness Description: Typical robotic actuation\
581
+ \ units have a motor and its transmission system aligned with the axis of the\
582
+ \ joint they actuate, resulting in large lateral encumbrance of the whole system.\n\
583
+ \nThis solution aims at solving this problem and allows to locate the last transmission\
584
+ \ stage at an arbitrary distance from the axis of the screw.\n\nTech Features:\
585
+ \ This invention consists in a class of actuation systems which is specifically\
586
+ \ designed to minimize the lateral encumbrance of the exoskeletal system to maximize\
587
+ \ its practical usability. Its core components are a motor coupled to a leadscrew\
588
+ \ or ballscrew system and a further (arbitrary) transmission system to connect\
589
+ \ the nut of the screw based transmission to the output wheel.\n\nFurthermore,\
590
+ \ this solution allows to displace the location of the screw with respect to the\
591
+ \ final transmission stage, allowing the adaptation of the location of the different\
592
+ \ stages of the system without influencing the overall behavior.The designed anti-blockage\
593
+ \ system guarantees proper functioning of each mechanical component and high back-drivability\
594
+ \ of the overall system even for high overall gearing.\n\nApplications: Robotics.\n\
595
+ \nAdvantages: Minimal lateral encumbrance; High gearing; Highly Back-drivable;\
596
+ \ Torque estimation through current measurement; Capability of dislocating the\
597
+ \ screw from the point of application of force."
598
+ - "Fluid filtering device based on a rotating valve pressure exchanger. Title: Fluid\
599
+ \ filtering device based on a rotating valve pressure exchanger\n\nAbstract: \n\
600
+ \nBusiness Description: The invention consists of an high pressure reciprocating\
601
+ \ pumping system equipped with an energy recovery device, intended for reverse\
602
+ \ osmosis plants. The patented device, descending by previous ENEA's patents,\
603
+ \ permits to realize water treatment plants with low energy consumption.\n\nTech\
604
+ \ Features: During ENEA's research in absorption refrigeration machines, have\
605
+ \ been patented several reciprocating pumping systems based on oil gear pumps\
606
+ \ coupled with an original self-adjusting rotating valve. Such a coupling, instead\
607
+ \ of the usual piston-rod-crank kinematic, has permitted to realize low cost and\
608
+ \ high reliability high pressure reciprocating pumps, called \"VR pumps\". Basing\
609
+ \ on the original know-how developed for the technology, a new invention has been\
610
+ \ patented, which consists to link a high pressure primary pump, VR type or not,\
611
+ \ to a second VR pump acted by the pressure of the waste water coming from the\
612
+ \ desalination plant, instead by pressurized oil as usual. Such second VR pump\
613
+ \ acts on the desalination circuit like an innovative pressure exchanger.\n\n\
614
+ Applications: Reverse osmosis desalination plants; Microfiltration desalination\
615
+ \ plants.\n\nAdvantages: Low cost; Easy scale up; Suitable for small plants; Low\
616
+ \ specific energy consumption."
617
+ - "Mechanical towing for automatic vehicle convoys. Title: Mechanical towing for\
618
+ \ automatic vehicle convoys\n\nAbstract: \n\nBusiness Description: The patent\
619
+ \ allows a mechanical coupling between vehicles in order to guarantee the safety\
620
+ \ and operation of a convoy of even 10 vehicles capable of circulating as if it\
621
+ \ were a single vehicle. This technology therefore allows a single driver in the\
622
+ \ head vehicle to automatically control the convoy of vehicles connected, revolutionizing\
623
+ \ the local transport systems with transport services that would otherwise not\
624
+ \ be feasible.\n\nTech Features: The patented mobile mechanical connection couples\
625
+ \ two vehicles and transforms them into a convoy that moves like a single unit.\
626
+ \ By means of an automatic guidance system, the coupled vehicles are able to synchronize\
627
+ \ automatically with the movements of the head vehicle, ensuring compliance with\
628
+ \ the trajectory of the head vehicle. The mechanical connection also acts as a\
629
+ \ safeguard in case of malfunctions and motion \"harmonizer\". This system allows\
630
+ \ single vehicles to form convoys driven by the driver only on the leading vehicle:\
631
+ \ free-flow car-sharing vehicles can be relocated from areas where they would\
632
+ \ be stationed for a long time to areas where there is demand, forming a convoy\
633
+ \ also of 10 vehicles and moving them with one driver; buses can be extended for\
634
+ \ the central sections of the transport lines with the highest demand without\
635
+ \ increasing the drivers; transport systems can be created in which small buses\
636
+ \ gather people on call in the suburbs and form a single convoy that crosses the\
637
+ \ center with a single driver. The patented coupling mechanism allows passenger\
638
+ \ transport companies to re-organize their services by means of convoys of vehicles\
639
+ \ which combine the maintenance of the optimum passenger capacity and the social\
640
+ \ requirements imposed by pandemic of COVID-19 and the containment measures taken.\n\
641
+ \nApplications: Car-sharing services; Logistics and goods distribution companies;\
642
+ \ Public and private transport with variable capacity.\n\nAdvantages: Certified\
643
+ \ transport system; Responsible automation; Flexible and client-oriented transport\
644
+ \ system; Efficiency of transport and sharing systems; Management cost reduction,\
645
+ \ creating capillary and high capacity services."
646
  pipeline_tag: sentence-similarity
647
  library_name: sentence-transformers
648
+ metrics:
649
+ - cosine_accuracy@1
650
+ - cosine_accuracy@3
651
+ - cosine_accuracy@5
652
+ - cosine_accuracy@10
653
+ - cosine_precision@1
654
+ - cosine_precision@3
655
+ - cosine_precision@5
656
+ - cosine_precision@10
657
+ - cosine_recall@1
658
+ - cosine_recall@3
659
+ - cosine_recall@5
660
+ - cosine_recall@10
661
+ - cosine_ndcg@10
662
+ - cosine_mrr@10
663
+ - cosine_map@100
664
+ model-index:
665
+ - name: SentenceTransformer based on AI-Growth-Lab/PatentSBERTa
666
+ results:
667
+ - task:
668
+ type: information-retrieval
669
+ name: Information Retrieval
670
+ dataset:
671
+ name: mnrl val
672
+ type: mnrl-val
673
+ metrics:
674
+ - type: cosine_accuracy@1
675
+ value: 0.039440203562340966
676
+ name: Cosine Accuracy@1
677
+ - type: cosine_accuracy@3
678
+ value: 0.08778625954198473
679
+ name: Cosine Accuracy@3
680
+ - type: cosine_accuracy@5
681
+ value: 0.14122137404580154
682
+ name: Cosine Accuracy@5
683
+ - type: cosine_accuracy@10
684
+ value: 0.2035623409669211
685
+ name: Cosine Accuracy@10
686
+ - type: cosine_precision@1
687
+ value: 0.039440203562340966
688
+ name: Cosine Precision@1
689
+ - type: cosine_precision@3
690
+ value: 0.029262086513994912
691
+ name: Cosine Precision@3
692
+ - type: cosine_precision@5
693
+ value: 0.028244274809160308
694
+ name: Cosine Precision@5
695
+ - type: cosine_precision@10
696
+ value: 0.020356234096692113
697
+ name: Cosine Precision@10
698
+ - type: cosine_recall@1
699
+ value: 0.039440203562340966
700
+ name: Cosine Recall@1
701
+ - type: cosine_recall@3
702
+ value: 0.08778625954198473
703
+ name: Cosine Recall@3
704
+ - type: cosine_recall@5
705
+ value: 0.14122137404580154
706
+ name: Cosine Recall@5
707
+ - type: cosine_recall@10
708
+ value: 0.2035623409669211
709
+ name: Cosine Recall@10
710
+ - type: cosine_ndcg@10
711
+ value: 0.10853541428278174
712
+ name: Cosine Ndcg@10
713
+ - type: cosine_mrr@10
714
+ value: 0.07969071852659648
715
+ name: Cosine Mrr@10
716
+ - type: cosine_map@100
717
+ value: 0.08979593108601017
718
+ name: Cosine Map@100
719
  ---
720
 
721
  # SentenceTransformer based on AI-Growth-Lab/PatentSBERTa
 
767
  model = SentenceTransformer("sentence_transformers_model_id")
768
  # Run inference
769
  sentences = [
770
+ 'High reduction and backdrivability ACTUATion SYSTEM with minimal lateral encumbrance. Title: High reduction and backdrivability ACTUATion SYSTEM with minimal lateral encumbrance\n\nAbstract: \n\nBusiness Description: Typical robotic actuation units have a motor and its transmission system aligned with the axis of the joint they actuate, resulting in large lateral encumbrance of the whole system.\n\nThis solution aims at solving this problem and allows to locate the last transmission stage at an arbitrary distance from the axis of the screw.\n\nTech Features: This invention consists in a class of actuation systems which is specifically designed to minimize the lateral encumbrance of the exoskeletal system to maximize its practical usability. Its core components are a motor coupled to a leadscrew or ballscrew system and a further (arbitrary) transmission system to connect the nut of the screw based transmission to the output wheel.\n\nFurthermore, this solution allows to displace the location of the screw with respect to the final transmission stage, allowing the adaptation of the location of the different stages of the system without influencing the overall behavior.The designed anti-blockage system guarantees proper functioning of each mechanical component and high back-drivability of the overall system even for high overall gearing.\n\nApplications: Robotics.\n\nAdvantages: Minimal lateral encumbrance; High gearing; Highly Back-drivable; Torque estimation through current measurement; Capability of dislocating the screw from the point of application of force.',
771
+ 'High reduction and backdrivability ACTUATion SYSTEM with minimal lateral encumbrance. Title: High reduction and backdrivability ACTUATion SYSTEM with minimal lateral encumbrance\n\nAbstract: \n\nBusiness Description: Typical robotic actuation units have a motor and its transmission system aligned with the axis of the joint they actuate, resulting in large lateral encumbrance of the whole system.\n\nThis solution aims at solving this problem and allows to locate the last transmission stage at an arbitrary distance from the axis of the screw.\n\nTech Features: This invention consists in a class of actuation systems which is specifically designed to minimize the lateral encumbrance of the exoskeletal system to maximize its practical usability. Its core components are a motor coupled to a leadscrew or ballscrew system and a further (arbitrary) transmission system to connect the nut of the screw based transmission to the output wheel.\n\nFurthermore, this solution allows to displace the location of the screw with respect to the final transmission stage, allowing the adaptation of the location of the different stages of the system without influencing the overall behavior.The designed anti-blockage system guarantees proper functioning of each mechanical component and high back-drivability of the overall system even for high overall gearing.\n\nApplications: Robotics.\n\nAdvantages: Minimal lateral encumbrance; High gearing; Highly Back-drivable; Torque estimation through current measurement; Capability of dislocating the screw from the point of application of force.',
772
+ 'Mechanical towing for automatic vehicle convoys. Title: Mechanical towing for automatic vehicle convoys\n\nAbstract: \n\nBusiness Description: The patent allows a mechanical coupling between vehicles in order to guarantee the safety and operation of a convoy of even 10 vehicles capable of circulating as if it were a single vehicle. This technology therefore allows a single driver in the head vehicle to automatically control the convoy of vehicles connected, revolutionizing the local transport systems with transport services that would otherwise not be feasible.\n\nTech Features: The patented mobile mechanical connection couples two vehicles and transforms them into a convoy that moves like a single unit. By means of an automatic guidance system, the coupled vehicles are able to synchronize automatically with the movements of the head vehicle, ensuring compliance with the trajectory of the head vehicle. The mechanical connection also acts as a safeguard in case of malfunctions and motion "harmonizer". This system allows single vehicles to form convoys driven by the driver only on the leading vehicle: free-flow car-sharing vehicles can be relocated from areas where they would be stationed for a long time to areas where there is demand, forming a convoy also of 10 vehicles and moving them with one driver; buses can be extended for the central sections of the transport lines with the highest demand without increasing the drivers; transport systems can be created in which small buses gather people on call in the suburbs and form a single convoy that crosses the center with a single driver. The patented coupling mechanism allows passenger transport companies to re-organize their services by means of convoys of vehicles which combine the maintenance of the optimum passenger capacity and the social requirements imposed by pandemic of COVID-19 and the containment measures taken.\n\nApplications: Car-sharing services; Logistics and goods distribution companies; Public and private transport with variable capacity.\n\nAdvantages: Certified transport system; Responsible automation; Flexible and client-oriented transport system; Efficiency of transport and sharing systems; Management cost reduction, creating capillary and high capacity services.',
773
  ]
774
  embeddings = model.encode(sentences)
775
  print(embeddings.shape)
 
778
  # Get the similarity scores for the embeddings
779
  similarities = model.similarity(embeddings, embeddings)
780
  print(similarities)
781
+ # tensor([[1.0000, 1.0000, 0.5066],
782
+ # [1.0000, 1.0000, 0.5066],
783
+ # [0.5066, 0.5066, 1.0000]])
784
  ```
785
 
786
  <!--
 
807
  *List how the model may foreseeably be misused and address what users ought not to do with the model.*
808
  -->
809
 
810
+ ## Evaluation
811
+
812
+ ### Metrics
813
+
814
+ #### Information Retrieval
815
+
816
+ * Dataset: `mnrl-val`
817
+ * Evaluated with [<code>InformationRetrievalEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator)
818
+
819
+ | Metric | Value |
820
+ |:--------------------|:-----------|
821
+ | cosine_accuracy@1 | 0.0394 |
822
+ | cosine_accuracy@3 | 0.0878 |
823
+ | cosine_accuracy@5 | 0.1412 |
824
+ | cosine_accuracy@10 | 0.2036 |
825
+ | cosine_precision@1 | 0.0394 |
826
+ | cosine_precision@3 | 0.0293 |
827
+ | cosine_precision@5 | 0.0282 |
828
+ | cosine_precision@10 | 0.0204 |
829
+ | cosine_recall@1 | 0.0394 |
830
+ | cosine_recall@3 | 0.0878 |
831
+ | cosine_recall@5 | 0.1412 |
832
+ | cosine_recall@10 | 0.2036 |
833
+ | **cosine_ndcg@10** | **0.1085** |
834
+ | cosine_mrr@10 | 0.0797 |
835
+ | cosine_map@100 | 0.0898 |
836
+
837
  <!--
838
  ## Bias, Risks and Limitations
839
 
 
852
 
853
  #### Unnamed Dataset
854
 
855
+ * Size: 6,288 training samples
856
  * Columns: <code>sentence_0</code> and <code>sentence_1</code>
857
  * Approximate statistics based on the first 1000 samples:
858
+ | | sentence_0 | sentence_1 |
859
+ |:--------|:---------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|
860
+ | type | string | string |
861
+ | details | <ul><li>min: 4 tokens</li><li>mean: 12.3 tokens</li><li>max: 35 tokens</li></ul> | <ul><li>min: 8 tokens</li><li>mean: 12.08 tokens</li><li>max: 36 tokens</li></ul> |
862
  * Samples:
863
+ | sentence_0 | sentence_1 |
864
+ |:--------------------------------------------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------|
865
+ | <code>Epigenetic regulation system for the control of target gene expression</code> | <code>Applicant/Organization: FONDAZIONE ISTITUTO ITALIANO DI TECNOLOGIA</code> |
866
+ | <code>System integrating a membrane humidifier and an adsorption-based storage for polymer membrane hydrogen fuel cell applications.</code> | <code>Technical Classification: H01M</code> |
867
+ | <code>Superconducting bipolar thermoelectric memory</code> | <code>Technical Classification: G11C_11</code> |
868
  * Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
869
  ```json
870
  {
 
877
  ### Training Hyperparameters
878
  #### Non-Default Hyperparameters
879
 
880
+ - `eval_strategy`: steps
881
  - `per_device_train_batch_size`: 12
882
  - `per_device_eval_batch_size`: 12
883
+ - `num_train_epochs`: 10
884
  - `multi_dataset_batch_sampler`: round_robin
885
 
886
  #### All Hyperparameters
887
  <details><summary>Click to expand</summary>
888
 
889
+ - `do_predict`: False
890
+ - `eval_strategy`: steps
891
+ - `prediction_loss_only`: True
892
  - `per_device_train_batch_size`: 12
893
+ - `per_device_eval_batch_size`: 12
894
+ - `gradient_accumulation_steps`: 1
895
+ - `eval_accumulation_steps`: None
896
+ - `torch_empty_cache_steps`: None
897
  - `learning_rate`: 5e-05
 
 
 
 
 
898
  - `weight_decay`: 0.0
899
  - `adam_beta1`: 0.9
900
  - `adam_beta2`: 0.999
901
  - `adam_epsilon`: 1e-08
 
 
 
902
  - `max_grad_norm`: 1
903
+ - `num_train_epochs`: 10
904
+ - `max_steps`: -1
905
+ - `lr_scheduler_type`: linear
906
+ - `lr_scheduler_kwargs`: None
907
+ - `warmup_ratio`: None
908
+ - `warmup_steps`: 0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
909
  - `log_level`: passive
910
  - `log_level_replica`: warning
911
+ - `log_on_each_node`: True
912
+ - `logging_nan_inf_filter`: True
 
 
 
 
 
 
 
 
 
 
 
 
913
  - `enable_jit_checkpoint`: False
914
+ - `save_on_each_node`: False
915
+ - `save_only_model`: False
 
 
 
 
 
 
916
  - `restore_callback_states_from_checkpoint`: False
917
+ - `use_cpu`: False
918
  - `seed`: 42
919
  - `data_seed`: None
920
+ - `bf16`: False
921
+ - `fp16`: False
922
+ - `bf16_full_eval`: False
923
+ - `fp16_full_eval`: False
924
+ - `tf32`: None
925
+ - `local_rank`: -1
926
+ - `ddp_backend`: None
927
+ - `debug`: []
928
  - `dataloader_drop_last`: False
929
  - `dataloader_num_workers`: 0
 
 
930
  - `dataloader_prefetch_factor`: None
931
+ - `disable_tqdm`: False
932
  - `remove_unused_columns`: True
933
  - `label_names`: None
934
+ - `load_best_model_at_end`: False
935
+ - `ignore_data_skip`: False
936
+ - `fsdp`: []
937
+ - `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
938
+ - `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
939
+ - `parallelism_config`: None
940
+ - `deepspeed`: None
941
+ - `label_smoothing_factor`: 0.0
942
+ - `optim`: adamw_torch_fused
943
+ - `optim_args`: None
944
+ - `group_by_length`: False
945
  - `length_column_name`: length
946
+ - `project`: huggingface
947
+ - `trackio_space_id`: trackio
948
  - `ddp_find_unused_parameters`: None
949
  - `ddp_bucket_cap_mb`: None
950
  - `ddp_broadcast_buffers`: False
951
+ - `dataloader_pin_memory`: True
952
+ - `dataloader_persistent_workers`: False
 
 
 
 
953
  - `skip_memory_metrics`: True
954
+ - `push_to_hub`: False
955
  - `resume_from_checkpoint`: None
956
+ - `hub_model_id`: None
957
+ - `hub_strategy`: every_save
958
+ - `hub_private_repo`: None
959
+ - `hub_always_push`: False
960
+ - `hub_revision`: None
961
+ - `gradient_checkpointing`: False
962
+ - `gradient_checkpointing_kwargs`: None
963
+ - `include_for_metrics`: []
964
+ - `eval_do_concat_batches`: True
965
+ - `auto_find_batch_size`: False
966
+ - `full_determinism`: False
967
+ - `ddp_timeout`: 1800
968
+ - `torch_compile`: False
969
+ - `torch_compile_backend`: None
970
+ - `torch_compile_mode`: None
971
+ - `include_num_input_tokens_seen`: no
972
+ - `neftune_noise_alpha`: None
973
+ - `optim_target_modules`: None
974
+ - `batch_eval_metrics`: False
975
+ - `eval_on_start`: False
976
+ - `use_liger_kernel`: False
977
+ - `liger_kernel_config`: None
978
+ - `eval_use_gather_object`: False
979
+ - `average_tokens_across_devices`: True
980
+ - `use_cache`: False
981
  - `prompts`: None
982
  - `batch_sampler`: batch_sampler
983
  - `multi_dataset_batch_sampler`: round_robin
 
987
  </details>
988
 
989
  ### Training Logs
990
+ | Epoch | Step | Training Loss | mnrl-val_cosine_ndcg@10 |
991
+ |:------:|:----:|:-------------:|:-----------------------:|
992
+ | 1.6835 | 500 | 0.0001 | - |
993
+ | 0.0954 | 50 | - | 0.0077 |
994
+ | 0.1908 | 100 | - | 0.0105 |
995
+ | 0.2863 | 150 | - | 0.0147 |
996
+ | 0.3817 | 200 | - | 0.0205 |
997
+ | 0.4771 | 250 | - | 0.0240 |
998
+ | 0.5725 | 300 | - | 0.0324 |
999
+ | 0.6679 | 350 | - | 0.0348 |
1000
+ | 0.7634 | 400 | - | 0.0332 |
1001
+ | 0.8588 | 450 | - | 0.0445 |
1002
+ | 0.9542 | 500 | 2.3022 | 0.0491 |
1003
+ | 1.0 | 524 | - | 0.0473 |
1004
+ | 1.0496 | 550 | - | 0.0479 |
1005
+ | 1.1450 | 600 | - | 0.0491 |
1006
+ | 1.2405 | 650 | - | 0.0466 |
1007
+ | 1.3359 | 700 | - | 0.0593 |
1008
+ | 1.4313 | 750 | - | 0.0547 |
1009
+ | 1.5267 | 800 | - | 0.0516 |
1010
+ | 1.6221 | 850 | - | 0.0596 |
1011
+ | 1.7176 | 900 | - | 0.0596 |
1012
+ | 1.8130 | 950 | - | 0.0681 |
1013
+ | 1.9084 | 1000 | 1.9086 | 0.0667 |
1014
+ | 2.0 | 1048 | - | 0.0666 |
1015
+ | 2.0038 | 1050 | - | 0.0686 |
1016
+ | 2.0992 | 1100 | - | 0.0732 |
1017
+ | 2.1947 | 1150 | - | 0.0686 |
1018
+ | 2.2901 | 1200 | - | 0.0772 |
1019
+ | 2.3855 | 1250 | - | 0.0752 |
1020
+ | 2.4809 | 1300 | - | 0.0803 |
1021
+ | 2.5763 | 1350 | - | 0.0721 |
1022
+ | 2.6718 | 1400 | - | 0.0779 |
1023
+ | 2.7672 | 1450 | - | 0.0745 |
1024
+ | 2.8626 | 1500 | 1.6854 | 0.0866 |
1025
+ | 2.9580 | 1550 | - | 0.0837 |
1026
+ | 3.0 | 1572 | - | 0.0788 |
1027
+ | 3.0534 | 1600 | - | 0.0752 |
1028
+ | 3.1489 | 1650 | - | 0.0788 |
1029
+ | 3.2443 | 1700 | - | 0.0845 |
1030
+ | 3.3397 | 1750 | - | 0.0898 |
1031
+ | 3.4351 | 1800 | - | 0.0920 |
1032
+ | 3.5305 | 1850 | - | 0.0877 |
1033
+ | 3.6260 | 1900 | - | 0.0926 |
1034
+ | 3.7214 | 1950 | - | 0.0858 |
1035
+ | 3.8168 | 2000 | 1.5140 | 0.0889 |
1036
+ | 3.9122 | 2050 | - | 0.0882 |
1037
+ | 4.0 | 2096 | - | 0.0848 |
1038
+ | 4.0076 | 2100 | - | 0.0828 |
1039
+ | 4.1031 | 2150 | - | 0.0910 |
1040
+ | 4.1985 | 2200 | - | 0.0928 |
1041
+ | 4.2939 | 2250 | - | 0.0913 |
1042
+ | 4.3893 | 2300 | - | 0.0923 |
1043
+ | 4.4847 | 2350 | - | 0.0888 |
1044
+ | 4.5802 | 2400 | - | 0.0882 |
1045
+ | 4.6756 | 2450 | - | 0.0987 |
1046
+ | 4.7710 | 2500 | 1.3415 | 0.0954 |
1047
+ | 4.8664 | 2550 | - | 0.0911 |
1048
+ | 4.9618 | 2600 | - | 0.0932 |
1049
+ | 5.0 | 2620 | - | 0.0887 |
1050
+ | 5.0573 | 2650 | - | 0.0952 |
1051
+ | 5.1527 | 2700 | - | 0.0954 |
1052
+ | 5.2481 | 2750 | - | 0.0972 |
1053
+ | 5.3435 | 2800 | - | 0.0957 |
1054
+ | 5.4389 | 2850 | - | 0.0999 |
1055
+ | 5.5344 | 2900 | - | 0.0964 |
1056
+ | 5.6298 | 2950 | - | 0.0980 |
1057
+ | 5.7252 | 3000 | 1.2411 | 0.0959 |
1058
+ | 5.8206 | 3050 | - | 0.0943 |
1059
+ | 5.9160 | 3100 | - | 0.0963 |
1060
+ | 6.0 | 3144 | - | 0.0914 |
1061
+ | 6.0115 | 3150 | - | 0.0915 |
1062
+ | 6.1069 | 3200 | - | 0.0974 |
1063
+ | 6.2023 | 3250 | - | 0.1019 |
1064
+ | 6.2977 | 3300 | - | 0.1014 |
1065
+ | 6.3931 | 3350 | - | 0.1037 |
1066
+ | 6.4885 | 3400 | - | 0.0987 |
1067
+ | 6.5840 | 3450 | - | 0.1010 |
1068
+ | 6.6794 | 3500 | 1.1304 | 0.1064 |
1069
+ | 6.7748 | 3550 | - | 0.1085 |
1070
 
1071
 
1072
  ### Framework Versions
1073
+ - Python: 3.12.12
1074
  - Sentence Transformers: 5.2.3
1075
+ - Transformers: 5.0.0
1076
+ - PyTorch: 2.10.0+cu128
1077
  - Accelerate: 1.13.0
1078
+ - Datasets: 4.0.0
1079
  - Tokenizers: 0.22.2
1080
 
1081
  ## Citation
config.json CHANGED
@@ -19,6 +19,6 @@
19
  "pad_token_id": 1,
20
  "relative_attention_num_buckets": 32,
21
  "tie_word_embeddings": true,
22
- "transformers_version": "5.3.0",
23
  "vocab_size": 30527
24
  }
 
19
  "pad_token_id": 1,
20
  "relative_attention_num_buckets": 32,
21
  "tie_word_embeddings": true,
22
+ "transformers_version": "5.0.0",
23
  "vocab_size": 30527
24
  }
config_sentence_transformers.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "__version__": {
3
  "sentence_transformers": "5.2.3",
4
- "transformers": "5.3.0",
5
- "pytorch": "2.10.0"
6
  },
7
  "model_type": "SentenceTransformer",
8
  "prompts": {
 
1
  {
2
  "__version__": {
3
  "sentence_transformers": "5.2.3",
4
+ "transformers": "5.0.0",
5
+ "pytorch": "2.10.0+cu128"
6
  },
7
  "model_type": "SentenceTransformer",
8
  "prompts": {
eval/Information-Retrieval_evaluation_mnrl-val_results.csv ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ epoch,steps,cosine-Accuracy@1,cosine-Accuracy@3,cosine-Accuracy@5,cosine-Accuracy@10,cosine-Precision@1,cosine-Recall@1,cosine-Precision@3,cosine-Recall@3,cosine-Precision@5,cosine-Recall@5,cosine-Precision@10,cosine-Recall@10,cosine-MRR@10,cosine-NDCG@10,cosine-MAP@100
2
+ 1.0,524,0.01653944020356234,0.03435114503816794,0.054707379134860054,0.0916030534351145,0.01653944020356234,0.01653944020356234,0.011450381679389313,0.03435114503816794,0.010941475826972013,0.054707379134860054,0.009160305343511452,0.0916030534351145,0.03400329173229936,0.04729229138376638,0.0446768559401947
3
+ 2.0,1048,0.017811704834605598,0.04580152671755725,0.06997455470737914,0.13994910941475827,0.017811704834605598,0.017811704834605598,0.015267175572519083,0.04580152671755725,0.013994910941475827,0.06997455470737914,0.013994910941475827,0.13994910941475827,0.044775637142049365,0.0666122455698021,0.05538703001323881
4
+ 3.0,1572,0.02544529262086514,0.0648854961832061,0.09033078880407125,0.15776081424936386,0.02544529262086514,0.02544529262086514,0.021628498727735368,0.0648854961832061,0.018066157760814248,0.09033078880407125,0.015776081424936386,0.15776081424936386,0.05532937517670343,0.07883996504876586,0.06667857333341984
5
+ 4.0,2096,0.021628498727735368,0.06361323155216285,0.11450381679389313,0.17302798982188294,0.021628498727735368,0.021628498727735368,0.021204410517387615,0.06361323155216285,0.022900763358778626,0.11450381679389313,0.017302798982188297,0.17302798982188294,0.05804202512217781,0.08483066508935018,0.06894986491588889
6
+ 5.0,2620,0.02544529262086514,0.07251908396946564,0.11323155216284987,0.17302798982188294,0.02544529262086514,0.02544529262086514,0.024173027989821884,0.07251908396946564,0.022646310432569976,0.11323155216284987,0.017302798982188297,0.17302798982188294,0.06303919786744216,0.08869954655146077,0.07417256951827288
7
+ 6.0,3144,0.03307888040712468,0.07633587786259542,0.11577608142493638,0.1717557251908397,0.03307888040712468,0.03307888040712468,0.025445292620865135,0.07633587786259542,0.02315521628498728,0.11577608142493638,0.01717557251908397,0.1717557251908397,0.06702512217779401,0.09136737265094202,0.07886692480908095
8
+ 7.0,3668,0.026717557251908396,0.07251908396946564,0.12086513994910941,0.19592875318066158,0.026717557251908396,0.026717557251908396,0.024173027989821884,0.07251908396946564,0.024173027989821884,0.12086513994910941,0.01959287531806616,0.19592875318066158,0.06682519487863005,0.09675471303342417,0.0768702700581889
9
+ 8.0,4192,0.030534351145038167,0.08651399491094147,0.1297709923664122,0.19592875318066158,0.030534351145038167,0.030534351145038167,0.028837998303647156,0.08651399491094147,0.025954198473282446,0.1297709923664122,0.01959287531806616,0.19592875318066158,0.07224696070115927,0.10118524847275165,0.08262985042550873
10
+ 9.0,4716,0.039440203562340966,0.09033078880407125,0.12340966921119594,0.18702290076335878,0.039440203562340966,0.039440203562340966,0.030110262934690414,0.09033078880407125,0.024681933842239184,0.12340966921119594,0.01870229007633588,0.18702290076335878,0.07651460075124207,0.10229083539418897,0.08795518859079528
11
+ 10.0,5240,0.03816793893129771,0.08651399491094147,0.1272264631043257,0.20483460559796438,0.03816793893129771,0.03816793893129771,0.028837998303647156,0.08651399491094147,0.02544529262086514,0.1272264631043257,0.02048346055979644,0.20483460559796438,0.07706288622319164,0.10668355939940834,0.08701220734782011
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7124e7a341b5a27bb2703ada4533fd5cbeaab50189f7313fa499b4f2c132263f
3
  size 437967648
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:776b6de4d824f001cee8169d99b7e387718aa61f61203c7707f0c6cfcd599dc5
3
  size 437967648
tokenizer_config.json CHANGED
@@ -6,18 +6,11 @@
6
  "eos_token": "</s>",
7
  "is_local": false,
8
  "mask_token": "<mask>",
9
- "max_length": 512,
10
  "model_max_length": 512,
11
- "pad_to_multiple_of": null,
12
  "pad_token": "<pad>",
13
- "pad_token_type_id": 0,
14
- "padding_side": "right",
15
  "sep_token": "</s>",
16
- "stride": 0,
17
  "strip_accents": null,
18
  "tokenize_chinese_chars": true,
19
  "tokenizer_class": "MPNetTokenizer",
20
- "truncation_side": "right",
21
- "truncation_strategy": "longest_first",
22
  "unk_token": "[UNK]"
23
  }
 
6
  "eos_token": "</s>",
7
  "is_local": false,
8
  "mask_token": "<mask>",
 
9
  "model_max_length": 512,
 
10
  "pad_token": "<pad>",
 
 
11
  "sep_token": "</s>",
 
12
  "strip_accents": null,
13
  "tokenize_chinese_chars": true,
14
  "tokenizer_class": "MPNetTokenizer",
 
 
15
  "unk_token": "[UNK]"
16
  }