tmnt 0.7.44b20240125__py3-none-any.whl → 0.7.44b20240127__py3-none-any.whl

tmnt/estimator.py

@@ -1405,7 +1405,7 @@ class SeqBowEstimator(BaseEstimator):
                 "params": [
                     p for n, p in model.llm.named_parameters() if (n in decay_parameters and p.requires_grad)
                 ],
-                "weight_decay": 1e-5,
+                "weight_decay": 1e-3,
             },
             { "params": [
                 p for n, p in model.llm.named_parameters() if (n not in decay_parameters and p.requires_grad)
@@ -1452,6 +1452,7 @@ class SeqBowEstimator(BaseEstimator):
             if self.metric is not None:
                 self.metric.reset()
             model.train()
+            model.llm.train()
             
             for (batch_id, (data, aux_batch)) in enumerate(joint_loader):
                 # data_batch is either a 2-tuple of: (labeled, unlabeled)
@@ -1468,11 +1469,14 @@ class SeqBowEstimator(BaseEstimator):
                 update_loss_details(total_ls, elbo_ls, red_ls, label_ls)
                 if aux_batch is not None:
                     update_loss_details(total_ls_2, elbo_ls_2, red_ls_2, None)
+                    
+                #debug
 
                 if not accumulate or (batch_id + 1) % accumulate == 0:
-                    torch.nn.utils.clip_grad.clip_grad_value_(model.llm.parameters(), 1.0)
-                    lr_scheduler.step()
+                    #torch.nn.utils.clip_grad.clip_grad_value_(model.llm.parameters(), 1.0)
+                    optimizer.step()
                     dec_optimizer.step()
+                    lr_scheduler.step()
                     model.zero_grad()
                     step_num += 1
                 if (batch_id + 1) % (self.log_interval) == 0:
@@ -1593,7 +1597,7 @@ class SeqBowMetricEstimator(SeqBowEstimator):
     def _get_model(self):
         llm_base_model = get_llm_model(self.llm_model_name).to(self.device)
         model = MetricSeqBowVED(llm_base_model, self.latent_distribution, num_classes=self.n_labels, device=self.device, 
-                                vocab_size = len(self.vocabulary), use_pooling=(self.llm_model_name.startswith("sentence-transformers")),
+                                vocab_size = len(self.vocabulary), use_pooling=self.pool_encoder,
                                 dropout=self.classifier_dropout, entropy_loss_coef=self.entropy_loss_coef)
         return model
         

tmnt/modeling.py

@@ -595,10 +595,11 @@ class MetricSeqBowVED(BaseSeqBowVED):
         elbo = elbo1 + elbo2
         rec_loss = rec_loss1 + rec_loss2
         KL_loss = KL_loss1 + KL_loss2
-        z_mu1 = self.latent_distribution.get_mu_encoding(enc1)
+        z_mu1 = self.latent_distribution.get_mu_encoding(enc2)
         z_mu2 = self.latent_distribution.get_mu_encoding(enc2)
         redundancy_loss = entropy_loss1 + entropy_loss2 #self.get_redundancy_penalty()
         return elbo, rec_loss, KL_loss, redundancy_loss, z_mu1, z_mu2
+        #return elbo, rec_loss, KL_loss, redundancy_loss, enc1, enc2
 
 
 class GeneralizedSDMLLoss(_Loss):

{tmnt-0.7.44b20240125.dist-info → tmnt-0.7.44b20240127.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tmnt
-Version: 0.7.44b20240125
+Version: 0.7.44b20240127
 Summary: Topic modeling neural toolkit
 Home-page: https://github.com/mitre/tmnt.git
 Author: The MITRE Corporation

{tmnt-0.7.44b20240125.dist-info → tmnt-0.7.44b20240127.dist-info}/RECORD

@@ -2,10 +2,10 @@ tmnt/__init__.py,sha256=EPNq1H7UMyMewWT_zTGBaC7ZouvCywX_gMX4G1dtmvw,250
 tmnt/configuration.py,sha256=P8PEhzVPKO5xG0FrdTLRQ60OYWigbzPY-OSx_hzQlrY,10054
 tmnt/data_loading.py,sha256=B47kfq5nrpw2bHYT2qEv2tpCLT7EFwqD7ZDjsoBto_Q,18303
 tmnt/distribution.py,sha256=Pmyc5gwDd_-jP7vLVb0vdNQaSSvF1EuiTZEWg3KfmI8,10866
-tmnt/estimator.py,sha256=IIXjtKB09qUqL_lDiDbhd5IVsW7hLuCHo82fF27xp64,77942
+tmnt/estimator.py,sha256=kQZ42MfOBBZuF0TQVdd9vBlw101ZlXk77mlws2ZvAS4,78014
 tmnt/eval_npmi.py,sha256=ODRDMsBgDM__iCNEX399ck7bAhl7ydvgDqmpfR7Y-q4,5048
 tmnt/inference.py,sha256=Sw7GO7QiWVEtbPJKBjFB7AiKRmUOZbFZn3tCrsStzWw,17845
-tmnt/modeling.py,sha256=Q-CSN0oaftf6RhM3Y3zKk4xw1Wd_WeZmPexZy8nk2Nw,32947
+tmnt/modeling.py,sha256=-fvmbT-KXr8luhELnCAOyZ-DUbTUd65cKRNRaH49EKI,33016
 tmnt/preprocess/__init__.py,sha256=gwMejkQrnqKS05i0JVsUru2hDUR5jE1hKC10dL934GU,170
 tmnt/preprocess/tokenizer.py,sha256=-ZgowfbHrM040vbNTktZM_hdl6HDTqxSJ4mDAxq3dUs,14050
 tmnt/preprocess/vectorizer.py,sha256=RkdivqP76qAJDianV09lONad9NbfBVWLZgIbU_P1-zo,15796
@@ -17,9 +17,9 @@ tmnt/utils/ngram_helpers.py,sha256=VrIzou2oQHCLBLSWODDeikN3PYat1NqqvEeYQj_GhbA,1
 tmnt/utils/pubmed_utils.py,sha256=3sHwoun7vxb0GV-arhpXLMUbAZne0huAh9xQNy6H40E,1274
 tmnt/utils/random.py,sha256=qY75WG3peWoMh9pUyCPBEo6q8IvkF6VRjeb5CqJOBF8,327
 tmnt/utils/recalibrate.py,sha256=TmpB8An8bslICZ13UTJfIvr8VoqiSedtpHxec4n8CHk,1439
-tmnt-0.7.44b20240125.dist-info/LICENSE,sha256=qFZJrfJ7Zi4IXDiyiGVrHWic_l1h2tc36tI8Z7rK9bs,11356
-tmnt-0.7.44b20240125.dist-info/METADATA,sha256=0duXA_NTiacN4bKgC10fnqMdPeOfVEHqy9EDz7EqquU,1403
-tmnt-0.7.44b20240125.dist-info/NOTICE,sha256=p0kYIVAkReTFaGb4C-qPa7h5ztze6hGzOpjCMMbOipU,425
-tmnt-0.7.44b20240125.dist-info/WHEEL,sha256=oiQVh_5PnQM0E3gPdiz09WCNmwiHDMaGer_elqB3coM,92
-tmnt-0.7.44b20240125.dist-info/top_level.txt,sha256=RpYgUl187sXnqmiwKjZZdcDlHz2AALs6bGdUcukyd_E,5
-tmnt-0.7.44b20240125.dist-info/RECORD,,
+tmnt-0.7.44b20240127.dist-info/LICENSE,sha256=qFZJrfJ7Zi4IXDiyiGVrHWic_l1h2tc36tI8Z7rK9bs,11356
+tmnt-0.7.44b20240127.dist-info/METADATA,sha256=RNb_SRd6cyvKGKSJT1NKTDdjjVVUfhDXqRuFIxmy2dE,1403
+tmnt-0.7.44b20240127.dist-info/NOTICE,sha256=p0kYIVAkReTFaGb4C-qPa7h5ztze6hGzOpjCMMbOipU,425
+tmnt-0.7.44b20240127.dist-info/WHEEL,sha256=oiQVh_5PnQM0E3gPdiz09WCNmwiHDMaGer_elqB3coM,92
+tmnt-0.7.44b20240127.dist-info/top_level.txt,sha256=RpYgUl187sXnqmiwKjZZdcDlHz2AALs6bGdUcukyd_E,5
+tmnt-0.7.44b20240127.dist-info/RECORD,,