Modified validation and training for linktransformer model

.gitattributes

@@ -38,3 +38,4 @@ pytorch_model.bin filter=lfs diff=lfs merge=lfs -text
 model.safetensors filter=lfs diff=lfs merge=lfs -text
 .git/lfs/objects/21/3e/213e31b950d84bee8f77e75c3cd557932d9bc7dc76dcd03aba223c2e4f841a0f filter=lfs diff=lfs merge=lfs -text
 .git/lfs/objects/3f/07/3f07f428adbf03a3e7902665664f5dd6f342a6b6c7a2a47a7026772c48cb0e73 filter=lfs diff=lfs merge=lfs -text
+.git/lfs/objects/3d/42/3d42ebe4f358451d3fde5790fe6516dcd94dc0e5c0159c25ebfe679e11eab41d filter=lfs diff=lfs merge=lfs -text

LT_training_config.json

@@ -1,10 +1,10 @@
 {
   "model_save_dir": "models",
   "model_save_name": "linkage_un_data_es_fine_industry",
-  "opt_model_description": "This model was trained on a dataset prepared by linking product classifications from [UN stats](https://unstats.un.org/unsd/classifications/Econ). \n                                 This model is designed to link different products to their industrial classification (ISIC) - trained on variation brought on by product level correspondance. It was trained for 70 epochs using other defaults that can be found in the repo's LinkTransformer config file - LT_training_config.json \n  ",
+  "opt_model_description": "This model was trained on a dataset prepared by linking product classifications from [UN stats](https://unstats.un.org/unsd/classifications/Econ). \n                                 This model is designed to link different products to their industrial classification (ISIC) - trained on variation brought on by product level correspondance. It was trained for 30 epochs using other defaults that can be found in the repo's LinkTransformer config file - LT_training_config.json \n  ",
   "opt_model_lang": "es",
   "train_batch_size": 64,
-  "num_epochs": 70,
+  "num_epochs": 30,
   "warm_up_perc": 1,
   "learning_rate": 2e-05,
   "loss_type": "supcon",
@@ -20,8 +20,10 @@
   "eval_steps_perc": 0.5,
   "test_at_end": true,
   "save_val_test_pickles": true,
-  "val_query_prop": 0.5,
+  "val_query_prop": 0.2,
   "loss_params": {},
+  "lr": 2e-06,
+  "warmup_perc": 0.5,
   "eval_type": "retrieval",
   "training_dataset": "dataframe",
   "base_model_path": "hiiamsid/sentence_similarity_spanish_es",

README.md

@@ -23,7 +23,7 @@ This model has been fine-tuned on the model : hiiamsid/sentence_similarity_spani
 
 
 This model was trained on a dataset prepared by linking product classifications from [UN stats](https://unstats.un.org/unsd/classifications/Econ). 
-                                 This model is designed to link different products to their industrial classification (ISIC) - trained on variation brought on by product level correspondance. It was trained for 70 epochs using other defaults that can be found in the repo's LinkTransformer config file - LT_training_config.json 
+                                 This model is designed to link different products to their industrial classification (ISIC) - trained on variation brought on by product level correspondance. It was trained for 30 epochs using other defaults that can be found in the repo's LinkTransformer config file - LT_training_config.json 
   
 
 ## Usage (LinkTransformer)
@@ -109,7 +109,7 @@ The model was trained with the parameters:
 Parameters of the fit()-Method:
 ```
 {
-    "epochs": 70,
+    "epochs": 30,
     "evaluation_steps": 33,
     "evaluator": "sentence_transformers.evaluation.SequentialEvaluator.SequentialEvaluator",
     "max_grad_norm": 1,
@@ -119,7 +119,7 @@ Parameters of the fit()-Method:
     },
     "scheduler": "WarmupLinear",
     "steps_per_epoch": null,
-    "warmup_steps": 4550,
+    "warmup_steps": 1950,
     "weight_decay": 0.01
 }
 ```

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d42ebe4f358451d3fde5790fe6516dcd94dc0e5c0159c25ebfe679e11eab41d
+oid sha256:82e8a4b706199c597b03c69c5184163d6bded50042bbb9e85f3928b2f5be3818
 size 439425888