diff --git a/articles/get_entities.html b/articles/get_entities.html index a53ab881..e337d704 100644 --- a/articles/get_entities.html +++ b/articles/get_entities.html @@ -176,7 +176,7 @@

Generic Approach U the command again will not trigger a download.

 tagger_ner <- load_tagger_ner("ner")
-#> 2023-11-29 09:41:52,148 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP>
+#> 2023-11-29 12:37:10,539 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP>

Flair NLP operates under the PyTorch framework. As such, we can use the $to method to set the device for the Flair Python library. The flair_device(“cpu”) allows you to select whether to use the @@ -368,7 +368,7 @@

Batch Processing#> Processing batch 2 out of 2... print(batch_process_time) #> user system elapsed -#> 23.521 0.211 23.593 +#> 23.833 0.219 23.966 
 print(batch_process_results)
 #>               doc_id                          entity  tag text_id
diff --git a/articles/get_pos.html b/articles/get_pos.html
index 57e23fc0..b4ba2a9e 100644
--- a/articles/get_pos.html
+++ b/articles/get_pos.html
@@ -172,7 +172,7 @@ 
Generic Approach Using Pa
 Hugging Face.

 
 tagger_pos <- load_tagger_pos("pos")
-#> 2023-11-29 09:42:49,554 SequenceTagger predicts: Dictionary with 53 tags: <unk>, O, UH, ,, VBD, PRP, VB, PRP$, NN, RB, ., DT, JJ, VBP, VBG, IN, CD, NNS, NNP, WRB, VBZ, WDT, CC, TO, MD, VBN, WP, :, RP, EX, JJR, FW, XX, HYPH, POS, RBR, JJS, PDT, NNPS, RBS, AFX, WP$, -LRB-, -RRB-, ``, '', LS, $, SYM, ADD

+#> 2023-11-29 12:38:08,493 SequenceTagger predicts: Dictionary with 53 tags: <unk>, O, UH, ,, VBD, PRP, VB, PRP$, NN, RB, ., DT, JJ, VBP, VBG, IN, CD, NNS, NNP, WRB, VBZ, WDT, CC, TO, MD, VBN, WP, :, RP, EX, JJR, FW, XX, HYPH, POS, RBR, JJS, PDT, NNPS, RBS, AFX, WP$, -LRB-, -RRB-, ``, '', LS, $, SYM, ADD

Flair NLP operates under the PyTorch framework. As such, we can use the $to method to set the device for the Flair Python library. The flair_device(“cpu”) allows you to select whether to use the diff --git a/articles/highlight_text.html b/articles/highlight_text.html index 24c62550..0575950a 100644 --- a/articles/highlight_text.html +++ b/articles/highlight_text.html @@ -168,7 +168,7 @@

Create Text with Named Entitiesdata("uk_immigration") uk_immigration <- uk_immigration[30,] tagger_ner <- load_tagger_ner("ner") -#> 2023-11-29 09:43:31,686 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP> +#> 2023-11-29 12:38:51,404 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP> result <- get_entities(uk_immigration$text, tagger = tagger_ner, show.text_id = FALSE diff --git a/articles/quickstart.html b/articles/quickstart.html index df36b812..de6cf78b 100644 --- a/articles/quickstart.html +++ b/articles/quickstart.html @@ -370,7 +370,7 @@

# load the NER tagger Classifier <- flair_nn()$Classifier tagger <- Classifier$load('ner') -#> 2023-11-29 09:43:49,703 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP> +#> 2023-11-29 12:39:09,314 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP> # run NER over sentence tagger$predict(sentence) @@ -409,7 +409,7 @@

# load the NER tagger Classifier <- flair_nn()$Classifier tagger <- Classifier$load('pos') -#> 2023-11-29 09:43:50,275 SequenceTagger predicts: Dictionary with 53 tags: <unk>, O, UH, ,, VBD, PRP, VB, PRP$, NN, RB, ., DT, JJ, VBP, VBG, IN, CD, NNS, NNP, WRB, VBZ, WDT, CC, TO, MD, VBN, WP, :, RP, EX, JJR, FW, XX, HYPH, POS, RBR, JJS, PDT, NNPS, RBS, AFX, WP$, -LRB-, -RRB-, ``, '', LS, $, SYM, ADD +#> 2023-11-29 12:39:09,875 SequenceTagger predicts: Dictionary with 53 tags: <unk>, O, UH, ,, VBD, PRP, VB, PRP$, NN, RB, ., DT, JJ, VBP, VBG, IN, CD, NNS, NNP, WRB, VBZ, WDT, CC, TO, MD, VBN, WP, :, RP, EX, JJR, FW, XX, HYPH, POS, RBR, JJS, PDT, NNPS, RBS, AFX, WP$, -LRB-, -RRB-, ``, '', LS, $, SYM, ADD # run NER over sentence tagger$predict(sentence) @@ -524,7 +524,7 @@ 

 # initiate TransformerWordEmbeddings
 embedding <- flair_embeddings.TransformerDocumentEmbeddings('bert-base-uncased')
-#> 2023-11-29 09:43:54,925 Using long sentences for Document embeddings is only beneficial for cls_pooling types 'mean' and 'max
+#> 2023-11-29 12:39:14,535 Using long sentences for Document embeddings is only beneficial for cls_pooling types 'mean' and 'max
 
 # create a sentence
 sentence <- flair_data.Sentence('The grass is green .')
@@ -698,7 +698,7 @@ 

 library(flaiR)
 tagger_pos <- load_tagger_pos("pos-fast")
-#> 2023-11-29 09:43:58,053 SequenceTagger predicts: Dictionary with 53 tags: <unk>, O, UH, ,, VBD, PRP, VB, PRP$, NN, RB, ., DT, JJ, VBP, VBG, IN, CD, NNS, NNP, WRB, VBZ, WDT, CC, TO, MD, VBN, WP, :, RP, EX, JJR, FW, XX, HYPH, POS, RBR, JJS, PDT, NNPS, RBS, AFX, WP$, -LRB-, -RRB-, ``, '', LS, $, SYM, ADD
+#> 2023-11-29 12:39:17,586 SequenceTagger predicts: Dictionary with 53 tags: <unk>, O, UH, ,, VBD, PRP, VB, PRP$, NN, RB, ., DT, JJ, VBP, VBG, IN, CD, NNS, NNP, WRB, VBZ, WDT, CC, TO, MD, VBN, WP, :, RP, EX, JJR, FW, XX, HYPH, POS, RBR, JJS, PDT, NNPS, RBS, AFX, WP$, -LRB-, -RRB-, ``, '', LS, $, SYM, ADD 
 results <- get_pos(texts, doc_ids, tagger_pos)
 head(results, n = 10)
@@ -727,7 +727,7 @@ 

 library(flaiR)
 tagger_ner <- load_tagger_ner("ner")
-#> 2023-11-29 09:43:59,225 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP>
+#> 2023-11-29 12:39:18,790 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP> 
 results <- get_entities(texts, doc_ids, tagger_ner)
 head(results, n = 10)
diff --git a/articles/tutorial.html b/articles/tutorial.html
index 62a46587..2b7dfb00 100644
--- a/articles/tutorial.html
+++ b/articles/tutorial.html
@@ -621,7 +621,7 @@ 

 corpus <- Corpus(train=train, 
                  # dev=test,
                  test=test)
-#> 2023-11-29 09:45:11,080 No dev split found. Using 0% (i.e. 471 samples) of the train split as dev data
+#> 2023-11-29 12:40:29,096 No dev split found. Using 0% (i.e. 471 samples) of the train split as dev data
 sprintf("Corpus object sizes - Train: %d | Test: %d | Dev: %d", 
         length(corpus$train), 
         length(corpus$test),
@@ -659,7 +659,7 @@ 

 # load the NER tagger
 Classifier <- flair_nn()$Classifier
 tagger <- Classifier$load('ner')
-#> 2023-11-29 09:45:12,595 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP>
+#> 2023-11-29 12:40:30,777 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP>
 
 # run NER over sentence
 tagger$predict(sentence)
@@ -698,7 +698,7 @@

# load the NER tagger Classifier <- flair_nn()$Classifier tagger <- Classifier$load('pos') -#> 2023-11-29 09:45:13,472 SequenceTagger predicts: Dictionary with 53 tags: <unk>, O, UH, ,, VBD, PRP, VB, PRP$, NN, RB, ., DT, JJ, VBP, VBG, IN, CD, NNS, NNP, WRB, VBZ, WDT, CC, TO, MD, VBN, WP, :, RP, EX, JJR, FW, XX, HYPH, POS, RBR, JJS, PDT, NNPS, RBS, AFX, WP$, -LRB-, -RRB-, ``, '', LS, $, SYM, ADD +#> 2023-11-29 12:40:31,664 SequenceTagger predicts: Dictionary with 53 tags: <unk>, O, UH, ,, VBD, PRP, VB, PRP$, NN, RB, ., DT, JJ, VBP, VBG, IN, CD, NNS, NNP, WRB, VBZ, WDT, CC, TO, MD, VBN, WP, :, RP, EX, JJR, FW, XX, HYPH, POS, RBR, JJS, PDT, NNPS, RBS, AFX, WP$, -LRB-, -RRB-, ``, '', LS, $, SYM, ADD # run NER over sentence tagger$predict(sentence) @@ -1241,15 +1241,15 @@

Loading a Tagged Corpus
 # downsize to 0.05
 corpus = IMDB()
-#> 2023-11-29 09:45:22,978 Reading data from /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced
-#> 2023-11-29 09:45:22,978 Train: /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced/train.txt
-#> 2023-11-29 09:45:22,978 Dev: None
-#> 2023-11-29 09:45:22,978 Test: None
-#> 2023-11-29 09:45:23,578 No test split found. Using 0% (i.e. 5000 samples) of the train split as test data
-#> 2023-11-29 09:45:23,593 No dev split found. Using 0% (i.e. 4500 samples) of the train split as dev data
-#> 2023-11-29 09:45:23,593 Initialized corpus /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced (label type name is 'sentiment')
+#> 2023-11-29 12:40:41,193 Reading data from /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced
+#> 2023-11-29 12:40:41,193 Train: /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced/train.txt
+#> 2023-11-29 12:40:41,193 Dev: None
+#> 2023-11-29 12:40:41,193 Test: None
+#> 2023-11-29 12:40:41,794 No test split found. Using 0% (i.e. 5000 samples) of the train split as test data
+#> 2023-11-29 12:40:41,809 No dev split found. Using 0% (i.e. 4500 samples) of the train split as dev data
+#> 2023-11-29 12:40:41,809 Initialized corpus /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced (label type name is 'sentiment')
 corpus$downsample(0.05)
-#> <flair.datasets.document_classification.IMDB object at 0x2f99f1bd0>
+#> <flair.datasets.document_classification.IMDB object at 0x2f46f3670>

Print the sizes in the corpus object as follows - test: %d | train: %d | dev: %d”

 +#> 2023-11-29 12:40:41,915 Computing label dictionary. Progress: +#> 2023-11-29 12:40:45,705 Dictionary created for label 'sentiment' with 2 values: POSITIVE (seen 1014 times), NEGATIVE (seen 1011 times)
@@ -1345,8 +1345,8 @@

Training the Model# specifies how embeddings are stored in RAM, ie."cpu", "cuda", "gpu", "mps". # embeddings_storage_mode = "mps", max_epochs=10L) -#> 2023-11-29 09:45:29,229 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,229 Model: "TextClassifier( +#> 2023-11-29 12:40:47,456 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,457 Model: "TextClassifier( #> (embeddings): DocumentPoolEmbeddings( #> fine_tune_mode=none, pooling=mean #> (embeddings): StackedEmbeddings( @@ -1364,186 +1364,186 @@

Training the Model#> (weights): None #> (weight_tensor) None #> )" -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,230 Corpus: 2025 train + 225 dev + 250 test sentences -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,230 Train: 2025 sentences -#> 2023-11-29 09:45:29,230 (train_with_dev=False, train_with_test=False) -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,230 Training Params: -#> 2023-11-29 09:45:29,230 - learning_rate: "0.1" -#> 2023-11-29 09:45:29,230 - mini_batch_size: "32" -#> 2023-11-29 09:45:29,230 - max_epochs: "10" -#> 2023-11-29 09:45:29,230 - shuffle: "True" -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,230 Plugins: -#> 2023-11-29 09:45:29,230 - AnnealOnPlateau | patience: '3', anneal_factor: '0.5', min_learning_rate: '0.0001' -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,230 Final evaluation on model from best epoch (best-model.pt) -#> 2023-11-29 09:45:29,230 - metric: "('micro avg', 'f1-score')" -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,230 Computation: -#> 2023-11-29 09:45:29,230 - compute on device: cpu -#> 2023-11-29 09:45:29,230 - embedding storage: cpu -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,230 Model training base path: "classifier" -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,230 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:29,964 epoch 1 - iter 6/64 - loss 0.88400855 - time (sec): 0.73 - samples/sec: 261.81 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:30,639 epoch 1 - iter 12/64 - loss 0.88396302 - time (sec): 1.41 - samples/sec: 272.61 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:31,578 epoch 1 - iter 18/64 - loss 0.90135088 - time (sec): 2.35 - samples/sec: 245.34 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:32,297 epoch 1 - iter 24/64 - loss 0.91910574 - time (sec): 3.07 - samples/sec: 250.46 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:32,975 epoch 1 - iter 30/64 - loss 0.92947888 - time (sec): 3.74 - samples/sec: 256.36 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:33,859 epoch 1 - iter 36/64 - loss 0.91739042 - time (sec): 4.63 - samples/sec: 248.92 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:34,564 epoch 1 - iter 42/64 - loss 0.91689622 - time (sec): 5.33 - samples/sec: 252.01 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:35,265 epoch 1 - iter 48/64 - loss 0.91121411 - time (sec): 6.03 - samples/sec: 254.52 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:35,999 epoch 1 - iter 54/64 - loss 0.90482956 - time (sec): 6.77 - samples/sec: 255.31 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:36,902 epoch 1 - iter 60/64 - loss 0.90788143 - time (sec): 7.67 - samples/sec: 250.28 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:37,218 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:37,218 EPOCH 1 done: loss 0.9077 - lr: 0.100000 -#> 2023-11-29 09:45:38,373 DEV : loss 0.8828572034835815 - f1-score (micro avg) 0.4533 -#> 2023-11-29 09:45:38,979 - 0 epochs without improvement -#> 2023-11-29 09:45:38,980 saving best model -#> 2023-11-29 09:45:39,398 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:40,192 epoch 2 - iter 6/64 - loss 0.88782200 - time (sec): 0.79 - samples/sec: 241.98 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:40,879 epoch 2 - iter 12/64 - loss 0.89557258 - time (sec): 1.48 - samples/sec: 259.40 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:41,616 epoch 2 - iter 18/64 - loss 0.88284143 - time (sec): 2.22 - samples/sec: 259.75 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:42,521 epoch 2 - iter 24/64 - loss 0.86361855 - time (sec): 3.12 - samples/sec: 245.94 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:43,244 epoch 2 - iter 30/64 - loss 0.86423665 - time (sec): 3.85 - samples/sec: 249.66 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:43,970 epoch 2 - iter 36/64 - loss 0.85561348 - time (sec): 4.57 - samples/sec: 251.96 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:44,687 epoch 2 - iter 42/64 - loss 0.85215194 - time (sec): 5.29 - samples/sec: 254.13 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:45,569 epoch 2 - iter 48/64 - loss 0.85517522 - time (sec): 6.17 - samples/sec: 248.93 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:46,358 epoch 2 - iter 54/64 - loss 0.84677640 - time (sec): 6.96 - samples/sec: 248.28 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:47,097 epoch 2 - iter 60/64 - loss 0.85021446 - time (sec): 7.70 - samples/sec: 249.39 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:47,563 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:47,563 EPOCH 2 done: loss 0.8530 - lr: 0.100000 -#> 2023-11-29 09:45:48,557 DEV : loss 0.8783490061759949 - f1-score (micro avg) 0.4533 -#> 2023-11-29 09:45:49,193 - 0 epochs without improvement -#> 2023-11-29 09:45:49,194 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:49,916 epoch 3 - iter 6/64 - loss 0.87799916 - time (sec): 0.72 - samples/sec: 265.93 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:50,843 epoch 3 - iter 12/64 - loss 0.89608843 - time (sec): 1.65 - samples/sec: 232.80 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:51,572 epoch 3 - iter 18/64 - loss 0.90057748 - time (sec): 2.38 - samples/sec: 242.17 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:52,306 epoch 3 - iter 24/64 - loss 0.89696234 - time (sec): 3.11 - samples/sec: 246.75 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:53,218 epoch 3 - iter 30/64 - loss 0.89147007 - time (sec): 4.02 - samples/sec: 238.59 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:53,965 epoch 3 - iter 36/64 - loss 0.89204659 - time (sec): 4.77 - samples/sec: 241.47 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:54,667 epoch 3 - iter 42/64 - loss 0.87829229 - time (sec): 5.47 - samples/sec: 245.57 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:55,379 epoch 3 - iter 48/64 - loss 0.87632222 - time (sec): 6.19 - samples/sec: 248.34 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:56,098 epoch 3 - iter 54/64 - loss 0.87269292 - time (sec): 6.90 - samples/sec: 250.30 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:56,985 epoch 3 - iter 60/64 - loss 0.87970024 - time (sec): 7.79 - samples/sec: 246.44 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:45:57,307 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:57,307 EPOCH 3 done: loss 0.8762 - lr: 0.100000 -#> 2023-11-29 09:45:58,455 DEV : loss 0.8697565793991089 - f1-score (micro avg) 0.4533 -#> 2023-11-29 09:45:59,065 - 0 epochs without improvement -#> 2023-11-29 09:45:59,065 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:45:59,875 epoch 4 - iter 6/64 - loss 0.88036998 - time (sec): 0.81 - samples/sec: 237.15 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:00,627 epoch 4 - iter 12/64 - loss 0.84520512 - time (sec): 1.56 - samples/sec: 245.90 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:01,523 epoch 4 - iter 18/64 - loss 0.84692961 - time (sec): 2.46 - samples/sec: 234.38 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:02,233 epoch 4 - iter 24/64 - loss 0.84134499 - time (sec): 3.17 - samples/sec: 242.50 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:02,944 epoch 4 - iter 30/64 - loss 0.85828587 - time (sec): 3.88 - samples/sec: 247.51 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:03,661 epoch 4 - iter 36/64 - loss 0.85654225 - time (sec): 4.60 - samples/sec: 250.68 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:04,413 epoch 4 - iter 42/64 - loss 0.84629689 - time (sec): 5.35 - samples/sec: 251.32 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:05,317 epoch 4 - iter 48/64 - loss 0.84268748 - time (sec): 6.25 - samples/sec: 245.71 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:06,018 epoch 4 - iter 54/64 - loss 0.84196810 - time (sec): 6.95 - samples/sec: 248.57 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:06,762 epoch 4 - iter 60/64 - loss 0.85795100 - time (sec): 7.70 - samples/sec: 249.46 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:07,236 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:07,236 EPOCH 4 done: loss 0.8562 - lr: 0.100000 -#> 2023-11-29 09:46:08,223 DEV : loss 0.8513666391372681 - f1-score (micro avg) 0.4533 -#> 2023-11-29 09:46:08,860 - 0 epochs without improvement -#> 2023-11-29 09:46:08,860 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:09,622 epoch 5 - iter 6/64 - loss 0.87054925 - time (sec): 0.76 - samples/sec: 252.11 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:10,315 epoch 5 - iter 12/64 - loss 0.88675048 - time (sec): 1.45 - samples/sec: 264.04 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:11,253 epoch 5 - iter 18/64 - loss 0.85735505 - time (sec): 2.39 - samples/sec: 240.79 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:11,975 epoch 5 - iter 24/64 - loss 0.86416615 - time (sec): 3.11 - samples/sec: 246.61 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:12,717 epoch 5 - iter 30/64 - loss 0.85780198 - time (sec): 3.86 - samples/sec: 248.90 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:13,465 epoch 5 - iter 36/64 - loss 0.85106002 - time (sec): 4.60 - samples/sec: 250.17 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:14,353 epoch 5 - iter 42/64 - loss 0.84119070 - time (sec): 5.49 - samples/sec: 244.69 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:15,145 epoch 5 - iter 48/64 - loss 0.84184100 - time (sec): 6.28 - samples/sec: 244.40 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:15,861 epoch 5 - iter 54/64 - loss 0.83640844 - time (sec): 7.00 - samples/sec: 246.82 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:16,600 epoch 5 - iter 60/64 - loss 0.83129187 - time (sec): 7.74 - samples/sec: 248.08 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:17,078 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:17,079 EPOCH 5 done: loss 0.8262 - lr: 0.100000 -#> 2023-11-29 09:46:18,236 DEV : loss 0.8393897414207458 - f1-score (micro avg) 0.4578 -#> 2023-11-29 09:46:18,652 - 0 epochs without improvement -#> 2023-11-29 09:46:18,653 saving best model -#> 2023-11-29 09:46:19,006 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:19,813 epoch 6 - iter 6/64 - loss 0.90684560 - time (sec): 0.81 - samples/sec: 238.02 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:20,737 epoch 6 - iter 12/64 - loss 0.86297455 - time (sec): 1.73 - samples/sec: 221.92 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:21,506 epoch 6 - iter 18/64 - loss 0.84712324 - time (sec): 2.50 - samples/sec: 230.38 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:22,224 epoch 6 - iter 24/64 - loss 0.84480650 - time (sec): 3.22 - samples/sec: 238.68 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:23,091 epoch 6 - iter 30/64 - loss 0.84118373 - time (sec): 4.08 - samples/sec: 235.04 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:23,861 epoch 6 - iter 36/64 - loss 0.84887512 - time (sec): 4.86 - samples/sec: 237.27 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:24,546 epoch 6 - iter 42/64 - loss 0.85247641 - time (sec): 5.54 - samples/sec: 242.61 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:25,317 epoch 6 - iter 48/64 - loss 0.83503058 - time (sec): 6.31 - samples/sec: 243.38 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:26,260 epoch 6 - iter 54/64 - loss 0.83527769 - time (sec): 7.25 - samples/sec: 238.23 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:26,972 epoch 6 - iter 60/64 - loss 0.82645011 - time (sec): 7.97 - samples/sec: 241.03 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:27,261 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:27,261 EPOCH 6 done: loss 0.8187 - lr: 0.100000 -#> 2023-11-29 09:46:28,414 DEV : loss 0.697002649307251 - f1-score (micro avg) 0.5289 -#> 2023-11-29 09:46:29,017 - 0 epochs without improvement -#> 2023-11-29 09:46:29,018 saving best model -#> 2023-11-29 09:46:29,346 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:30,144 epoch 7 - iter 6/64 - loss 0.92862150 - time (sec): 0.80 - samples/sec: 240.83 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:30,835 epoch 7 - iter 12/64 - loss 0.88754721 - time (sec): 1.49 - samples/sec: 258.07 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:31,606 epoch 7 - iter 18/64 - loss 0.87636076 - time (sec): 2.26 - samples/sec: 254.97 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:32,542 epoch 7 - iter 24/64 - loss 0.86338259 - time (sec): 3.19 - samples/sec: 240.38 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:33,309 epoch 7 - iter 30/64 - loss 0.86797423 - time (sec): 3.96 - samples/sec: 242.26 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:34,053 epoch 7 - iter 36/64 - loss 0.85439281 - time (sec): 4.71 - samples/sec: 244.76 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:34,788 epoch 7 - iter 42/64 - loss 0.85024860 - time (sec): 5.44 - samples/sec: 247.01 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:35,647 epoch 7 - iter 48/64 - loss 0.83954957 - time (sec): 6.30 - samples/sec: 243.78 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:36,397 epoch 7 - iter 54/64 - loss 0.83340724 - time (sec): 7.05 - samples/sec: 245.08 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:37,109 epoch 7 - iter 60/64 - loss 0.83293155 - time (sec): 7.76 - samples/sec: 247.36 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:37,561 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:37,561 EPOCH 7 done: loss 0.8286 - lr: 0.100000 -#> 2023-11-29 09:46:38,547 DEV : loss 0.7402542233467102 - f1-score (micro avg) 0.5022 -#> 2023-11-29 09:46:39,182 - 1 epochs without improvement -#> 2023-11-29 09:46:39,183 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:39,967 epoch 8 - iter 6/64 - loss 0.72758570 - time (sec): 0.78 - samples/sec: 245.06 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:40,713 epoch 8 - iter 12/64 - loss 0.77498165 - time (sec): 1.53 - samples/sec: 251.09 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:41,576 epoch 8 - iter 18/64 - loss 0.78188934 - time (sec): 2.39 - samples/sec: 240.68 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:42,334 epoch 8 - iter 24/64 - loss 0.79232205 - time (sec): 3.15 - samples/sec: 243.76 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:43,067 epoch 8 - iter 30/64 - loss 0.78611903 - time (sec): 3.88 - samples/sec: 247.20 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:43,805 epoch 8 - iter 36/64 - loss 0.77288217 - time (sec): 4.62 - samples/sec: 249.25 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:44,705 epoch 8 - iter 42/64 - loss 0.77235644 - time (sec): 5.52 - samples/sec: 243.42 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:45,391 epoch 8 - iter 48/64 - loss 0.78255222 - time (sec): 6.21 - samples/sec: 247.43 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:46,135 epoch 8 - iter 54/64 - loss 0.78456129 - time (sec): 6.95 - samples/sec: 248.55 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:46,905 epoch 8 - iter 60/64 - loss 0.77978850 - time (sec): 7.72 - samples/sec: 248.65 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:47,357 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:47,357 EPOCH 8 done: loss 0.7780 - lr: 0.100000 -#> 2023-11-29 09:46:48,348 DEV : loss 0.7568300366401672 - f1-score (micro avg) 0.4978 -#> 2023-11-29 09:46:48,984 - 2 epochs without improvement -#> 2023-11-29 09:46:48,986 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:49,726 epoch 9 - iter 6/64 - loss 0.83446981 - time (sec): 0.74 - samples/sec: 259.52 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:50,655 epoch 9 - iter 12/64 - loss 0.83481617 - time (sec): 1.67 - samples/sec: 230.11 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:51,339 epoch 9 - iter 18/64 - loss 0.84183356 - time (sec): 2.35 - samples/sec: 244.80 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:52,052 epoch 9 - iter 24/64 - loss 0.82389081 - time (sec): 3.07 - samples/sec: 250.54 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:52,837 epoch 9 - iter 30/64 - loss 0.81802387 - time (sec): 3.85 - samples/sec: 249.32 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:53,727 epoch 9 - iter 36/64 - loss 0.81533981 - time (sec): 4.74 - samples/sec: 243.04 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:54,433 epoch 9 - iter 42/64 - loss 0.80514485 - time (sec): 5.45 - samples/sec: 246.75 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:55,196 epoch 9 - iter 48/64 - loss 0.81017195 - time (sec): 6.21 - samples/sec: 247.36 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:55,917 epoch 9 - iter 54/64 - loss 0.80359553 - time (sec): 6.93 - samples/sec: 249.35 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:56,798 epoch 9 - iter 60/64 - loss 0.79928778 - time (sec): 7.81 - samples/sec: 245.80 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:46:57,136 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:57,136 EPOCH 9 done: loss 0.7969 - lr: 0.100000 -#> 2023-11-29 09:46:58,275 DEV : loss 0.7745500206947327 - f1-score (micro avg) 0.4844 -#> 2023-11-29 09:46:58,881 - 3 epochs without improvement -#> 2023-11-29 09:46:58,882 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:46:59,709 epoch 10 - iter 6/64 - loss 0.74788894 - time (sec): 0.83 - samples/sec: 232.13 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:00,395 epoch 10 - iter 12/64 - loss 0.76720373 - time (sec): 1.51 - samples/sec: 253.81 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:01,152 epoch 10 - iter 18/64 - loss 0.79161676 - time (sec): 2.27 - samples/sec: 253.72 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:01,913 epoch 10 - iter 24/64 - loss 0.78399789 - time (sec): 3.03 - samples/sec: 253.41 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:02,789 epoch 10 - iter 30/64 - loss 0.78045449 - time (sec): 3.91 - samples/sec: 245.74 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:03,491 epoch 10 - iter 36/64 - loss 0.78069342 - time (sec): 4.61 - samples/sec: 249.96 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:04,246 epoch 10 - iter 42/64 - loss 0.76349049 - time (sec): 5.36 - samples/sec: 250.56 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:05,005 epoch 10 - iter 48/64 - loss 0.76165402 - time (sec): 6.12 - samples/sec: 250.88 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:05,888 epoch 10 - iter 54/64 - loss 0.77187297 - time (sec): 7.01 - samples/sec: 246.65 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:06,612 epoch 10 - iter 60/64 - loss 0.77289468 - time (sec): 7.73 - samples/sec: 248.38 - lr: 0.100000 - momentum: 0.000000 -#> 2023-11-29 09:47:07,105 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:47:07,105 EPOCH 10 done: loss 0.7810 - lr: 0.100000 -#> 2023-11-29 09:47:08,087 DEV : loss 0.7909061908721924 - f1-score (micro avg) 0.4889 -#> 2023-11-29 09:47:08,719 - 4 epochs without improvement (above 'patience')-> annealing learning_rate to [0.05] -#> 2023-11-29 09:47:09,052 ---------------------------------------------------------------------------------------------------- -#> 2023-11-29 09:47:09,053 Loading model from best epoch ... -#> 2023-11-29 09:47:10,226 +#> 2023-11-29 12:40:47,457 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,457 Corpus: 2025 train + 225 dev + 250 test sentences +#> 2023-11-29 12:40:47,457 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,457 Train: 2025 sentences +#> 2023-11-29 12:40:47,457 (train_with_dev=False, train_with_test=False) +#> 2023-11-29 12:40:47,457 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,457 Training Params: +#> 2023-11-29 12:40:47,457 - learning_rate: "0.1" +#> 2023-11-29 12:40:47,457 - mini_batch_size: "32" +#> 2023-11-29 12:40:47,457 - max_epochs: "10" +#> 2023-11-29 12:40:47,457 - shuffle: "True" +#> 2023-11-29 12:40:47,457 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,457 Plugins: +#> 2023-11-29 12:40:47,457 - AnnealOnPlateau | patience: '3', anneal_factor: '0.5', min_learning_rate: '0.0001' +#> 2023-11-29 12:40:47,457 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,457 Final evaluation on model from best epoch (best-model.pt) +#> 2023-11-29 12:40:47,457 - metric: "('micro avg', 'f1-score')" +#> 2023-11-29 12:40:47,457 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,457 Computation: +#> 2023-11-29 12:40:47,457 - compute on device: cpu +#> 2023-11-29 12:40:47,457 - embedding storage: cpu +#> 2023-11-29 12:40:47,457 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,458 Model training base path: "classifier" +#> 2023-11-29 12:40:47,458 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:47,458 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:48,196 epoch 1 - iter 6/64 - loss 0.88400855 - time (sec): 0.74 - samples/sec: 260.03 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:48,865 epoch 1 - iter 12/64 - loss 0.88396302 - time (sec): 1.41 - samples/sec: 272.95 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:49,794 epoch 1 - iter 18/64 - loss 0.90135088 - time (sec): 2.34 - samples/sec: 246.57 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:50,499 epoch 1 - iter 24/64 - loss 0.91910574 - time (sec): 3.04 - samples/sec: 252.53 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:51,178 epoch 1 - iter 30/64 - loss 0.92947888 - time (sec): 3.72 - samples/sec: 258.02 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:51,891 epoch 1 - iter 36/64 - loss 0.91739042 - time (sec): 4.43 - samples/sec: 259.88 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:52,736 epoch 1 - iter 42/64 - loss 0.91689622 - time (sec): 5.28 - samples/sec: 254.62 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:53,443 epoch 1 - iter 48/64 - loss 0.91121411 - time (sec): 5.98 - samples/sec: 256.65 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:54,158 epoch 1 - iter 54/64 - loss 0.90482956 - time (sec): 6.70 - samples/sec: 257.91 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:54,883 epoch 1 - iter 60/64 - loss 0.90788143 - time (sec): 7.43 - samples/sec: 258.58 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:55,338 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:55,338 EPOCH 1 done: loss 0.9077 - lr: 0.100000 +#> 2023-11-29 12:40:56,499 DEV : loss 0.8828572034835815 - f1-score (micro avg) 0.4533 +#> 2023-11-29 12:40:56,909 - 0 epochs without improvement +#> 2023-11-29 12:40:56,909 saving best model +#> 2023-11-29 12:40:57,319 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:40:58,255 epoch 2 - iter 6/64 - loss 0.88782200 - time (sec): 0.94 - samples/sec: 205.29 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:58,925 epoch 2 - iter 12/64 - loss 0.89557258 - time (sec): 1.61 - samples/sec: 239.21 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:40:59,655 epoch 2 - iter 18/64 - loss 0.88284143 - time (sec): 2.33 - samples/sec: 246.70 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:00,378 epoch 2 - iter 24/64 - loss 0.86361855 - time (sec): 3.06 - samples/sec: 251.16 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:01,102 epoch 2 - iter 30/64 - loss 0.86423665 - time (sec): 3.78 - samples/sec: 253.79 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:01,959 epoch 2 - iter 36/64 - loss 0.85561348 - time (sec): 4.64 - samples/sec: 248.29 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:02,665 epoch 2 - iter 42/64 - loss 0.85215194 - time (sec): 5.34 - samples/sec: 251.45 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:03,376 epoch 2 - iter 48/64 - loss 0.85517522 - time (sec): 6.06 - samples/sec: 253.61 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:04,313 epoch 2 - iter 54/64 - loss 0.84677640 - time (sec): 6.99 - samples/sec: 247.10 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:05,043 epoch 2 - iter 60/64 - loss 0.85021446 - time (sec): 7.72 - samples/sec: 248.61 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:05,336 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:05,336 EPOCH 2 done: loss 0.8530 - lr: 0.100000 +#> 2023-11-29 12:41:06,493 DEV : loss 0.8783490061759949 - f1-score (micro avg) 0.4533 +#> 2023-11-29 12:41:07,078 - 0 epochs without improvement +#> 2023-11-29 12:41:07,079 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:07,840 epoch 3 - iter 6/64 - loss 0.87799916 - time (sec): 0.76 - samples/sec: 252.28 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:08,584 epoch 3 - iter 12/64 - loss 0.89608843 - time (sec): 1.51 - samples/sec: 255.13 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:09,301 epoch 3 - iter 18/64 - loss 0.90057748 - time (sec): 2.22 - samples/sec: 259.30 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:10,179 epoch 3 - iter 24/64 - loss 0.89696234 - time (sec): 3.10 - samples/sec: 247.76 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:10,899 epoch 3 - iter 30/64 - loss 0.89147007 - time (sec): 3.82 - samples/sec: 251.29 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:11,631 epoch 3 - iter 36/64 - loss 0.89204659 - time (sec): 4.55 - samples/sec: 253.07 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:12,333 epoch 3 - iter 42/64 - loss 0.87829229 - time (sec): 5.25 - samples/sec: 255.83 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:13,193 epoch 3 - iter 48/64 - loss 0.87632222 - time (sec): 6.11 - samples/sec: 251.24 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:13,903 epoch 3 - iter 54/64 - loss 0.87269292 - time (sec): 6.82 - samples/sec: 253.23 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:14,606 epoch 3 - iter 60/64 - loss 0.87970024 - time (sec): 7.53 - samples/sec: 255.08 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:15,079 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:15,079 EPOCH 3 done: loss 0.8762 - lr: 0.100000 +#> 2023-11-29 12:41:16,056 DEV : loss 0.8697565793991089 - f1-score (micro avg) 0.4533 +#> 2023-11-29 12:41:16,672 - 0 epochs without improvement +#> 2023-11-29 12:41:16,673 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:17,430 epoch 4 - iter 6/64 - loss 0.88036998 - time (sec): 0.76 - samples/sec: 253.57 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:18,340 epoch 4 - iter 12/64 - loss 0.84520512 - time (sec): 1.67 - samples/sec: 230.34 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:19,052 epoch 4 - iter 18/64 - loss 0.84692961 - time (sec): 2.38 - samples/sec: 242.16 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:19,754 epoch 4 - iter 24/64 - loss 0.84134499 - time (sec): 3.08 - samples/sec: 249.26 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:20,453 epoch 4 - iter 30/64 - loss 0.85828587 - time (sec): 3.78 - samples/sec: 254.00 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:21,155 epoch 4 - iter 36/64 - loss 0.85654225 - time (sec): 4.48 - samples/sec: 257.04 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:22,052 epoch 4 - iter 42/64 - loss 0.84629689 - time (sec): 5.38 - samples/sec: 249.84 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:22,781 epoch 4 - iter 48/64 - loss 0.84268748 - time (sec): 6.11 - samples/sec: 251.49 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:23,468 epoch 4 - iter 54/64 - loss 0.84196810 - time (sec): 6.80 - samples/sec: 254.30 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:24,363 epoch 4 - iter 60/64 - loss 0.85795100 - time (sec): 7.69 - samples/sec: 249.69 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:24,660 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:24,660 EPOCH 4 done: loss 0.8562 - lr: 0.100000 +#> 2023-11-29 12:41:25,814 DEV : loss 0.8513666391372681 - f1-score (micro avg) 0.4533 +#> 2023-11-29 12:41:26,249 - 0 epochs without improvement +#> 2023-11-29 12:41:26,249 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:27,027 epoch 5 - iter 6/64 - loss 0.87054925 - time (sec): 0.78 - samples/sec: 246.94 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:27,872 epoch 5 - iter 12/64 - loss 0.88675048 - time (sec): 1.62 - samples/sec: 236.67 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:28,628 epoch 5 - iter 18/64 - loss 0.85735505 - time (sec): 2.38 - samples/sec: 242.23 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:29,335 epoch 5 - iter 24/64 - loss 0.86416615 - time (sec): 3.09 - samples/sec: 248.92 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:30,043 epoch 5 - iter 30/64 - loss 0.85780198 - time (sec): 3.79 - samples/sec: 253.08 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:30,931 epoch 5 - iter 36/64 - loss 0.85106002 - time (sec): 4.68 - samples/sec: 246.11 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:31,643 epoch 5 - iter 42/64 - loss 0.84119070 - time (sec): 5.39 - samples/sec: 249.19 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:32,399 epoch 5 - iter 48/64 - loss 0.84184100 - time (sec): 6.15 - samples/sec: 249.77 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:33,109 epoch 5 - iter 54/64 - loss 0.83640844 - time (sec): 6.86 - samples/sec: 251.91 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:34,002 epoch 5 - iter 60/64 - loss 0.83129187 - time (sec): 7.75 - samples/sec: 247.68 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:34,308 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:34,308 EPOCH 5 done: loss 0.8262 - lr: 0.100000 +#> 2023-11-29 12:41:35,445 DEV : loss 0.8393897414207458 - f1-score (micro avg) 0.4578 +#> 2023-11-29 12:41:36,037 - 0 epochs without improvement +#> 2023-11-29 12:41:36,038 saving best model +#> 2023-11-29 12:41:36,366 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:37,124 epoch 6 - iter 6/64 - loss 0.90684560 - time (sec): 0.76 - samples/sec: 253.32 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:37,873 epoch 6 - iter 12/64 - loss 0.86297455 - time (sec): 1.51 - samples/sec: 254.77 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:38,622 epoch 6 - iter 18/64 - loss 0.84712324 - time (sec): 2.26 - samples/sec: 255.33 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:39,331 epoch 6 - iter 24/64 - loss 0.84480650 - time (sec): 2.97 - samples/sec: 259.01 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:40,180 epoch 6 - iter 30/64 - loss 0.84118373 - time (sec): 3.81 - samples/sec: 251.72 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:40,935 epoch 6 - iter 36/64 - loss 0.84887512 - time (sec): 4.57 - samples/sec: 252.14 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:41,622 epoch 6 - iter 42/64 - loss 0.85247641 - time (sec): 5.26 - samples/sec: 255.69 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:42,378 epoch 6 - iter 48/64 - loss 0.83503058 - time (sec): 6.01 - samples/sec: 255.47 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:43,341 epoch 6 - iter 54/64 - loss 0.83527769 - time (sec): 6.98 - samples/sec: 247.72 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:44,042 epoch 6 - iter 60/64 - loss 0.82645011 - time (sec): 7.68 - samples/sec: 250.12 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:44,484 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:44,485 EPOCH 6 done: loss 0.8187 - lr: 0.100000 +#> 2023-11-29 12:41:45,456 DEV : loss 0.697002649307251 - f1-score (micro avg) 0.5289 +#> 2023-11-29 12:41:46,081 - 0 epochs without improvement +#> 2023-11-29 12:41:46,081 saving best model +#> 2023-11-29 12:41:46,397 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:47,185 epoch 7 - iter 6/64 - loss 0.92862150 - time (sec): 0.79 - samples/sec: 244.01 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:47,860 epoch 7 - iter 12/64 - loss 0.88754721 - time (sec): 1.46 - samples/sec: 262.63 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:48,617 epoch 7 - iter 18/64 - loss 0.87636076 - time (sec): 2.22 - samples/sec: 259.52 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:49,534 epoch 7 - iter 24/64 - loss 0.86338259 - time (sec): 3.14 - samples/sec: 244.90 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:50,286 epoch 7 - iter 30/64 - loss 0.86797423 - time (sec): 3.89 - samples/sec: 246.87 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:51,006 epoch 7 - iter 36/64 - loss 0.85439281 - time (sec): 4.61 - samples/sec: 249.96 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:51,730 epoch 7 - iter 42/64 - loss 0.85024860 - time (sec): 5.33 - samples/sec: 252.04 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:52,568 epoch 7 - iter 48/64 - loss 0.83954957 - time (sec): 6.17 - samples/sec: 248.93 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:53,309 epoch 7 - iter 54/64 - loss 0.83340724 - time (sec): 6.91 - samples/sec: 250.03 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:54,008 epoch 7 - iter 60/64 - loss 0.83293155 - time (sec): 7.61 - samples/sec: 252.30 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:54,448 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:54,448 EPOCH 7 done: loss 0.8286 - lr: 0.100000 +#> 2023-11-29 12:41:55,424 DEV : loss 0.7402542233467102 - f1-score (micro avg) 0.5022 +#> 2023-11-29 12:41:56,039 - 1 epochs without improvement +#> 2023-11-29 12:41:56,040 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:41:56,804 epoch 8 - iter 6/64 - loss 0.72758570 - time (sec): 0.76 - samples/sec: 251.29 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:57,547 epoch 8 - iter 12/64 - loss 0.77498165 - time (sec): 1.51 - samples/sec: 254.81 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:58,391 epoch 8 - iter 18/64 - loss 0.78188934 - time (sec): 2.35 - samples/sec: 245.02 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:59,124 epoch 8 - iter 24/64 - loss 0.79232205 - time (sec): 3.08 - samples/sec: 249.02 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:41:59,847 epoch 8 - iter 30/64 - loss 0.78611903 - time (sec): 3.81 - samples/sec: 252.17 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:00,597 epoch 8 - iter 36/64 - loss 0.77288217 - time (sec): 4.56 - samples/sec: 252.80 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:01,512 epoch 8 - iter 42/64 - loss 0.77235644 - time (sec): 5.47 - samples/sec: 245.62 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:02,221 epoch 8 - iter 48/64 - loss 0.78255222 - time (sec): 6.18 - samples/sec: 248.52 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:02,953 epoch 8 - iter 54/64 - loss 0.78456129 - time (sec): 6.91 - samples/sec: 249.97 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:03,717 epoch 8 - iter 60/64 - loss 0.77978850 - time (sec): 7.68 - samples/sec: 250.11 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:04,158 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:42:04,158 EPOCH 8 done: loss 0.7780 - lr: 0.100000 +#> 2023-11-29 12:42:05,135 DEV : loss 0.7568300366401672 - f1-score (micro avg) 0.4978 +#> 2023-11-29 12:42:05,746 - 2 epochs without improvement +#> 2023-11-29 12:42:05,747 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:42:06,639 epoch 9 - iter 6/64 - loss 0.83446981 - time (sec): 0.89 - samples/sec: 215.31 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:07,370 epoch 9 - iter 12/64 - loss 0.83481617 - time (sec): 1.62 - samples/sec: 236.69 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:08,059 epoch 9 - iter 18/64 - loss 0.84183356 - time (sec): 2.31 - samples/sec: 249.18 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:08,766 epoch 9 - iter 24/64 - loss 0.82389081 - time (sec): 3.02 - samples/sec: 254.42 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:09,698 epoch 9 - iter 30/64 - loss 0.81802387 - time (sec): 3.95 - samples/sec: 243.04 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:10,427 epoch 9 - iter 36/64 - loss 0.81533981 - time (sec): 4.68 - samples/sec: 246.20 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:11,116 epoch 9 - iter 42/64 - loss 0.80514485 - time (sec): 5.37 - samples/sec: 250.37 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:11,871 epoch 9 - iter 48/64 - loss 0.81017195 - time (sec): 6.12 - samples/sec: 250.85 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:12,583 epoch 9 - iter 54/64 - loss 0.80359553 - time (sec): 6.84 - samples/sec: 252.81 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:13,285 epoch 9 - iter 60/64 - loss 0.79928778 - time (sec): 7.54 - samples/sec: 254.75 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:13,782 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:42:13,782 EPOCH 9 done: loss 0.7969 - lr: 0.100000 +#> 2023-11-29 12:42:14,910 DEV : loss 0.7745500206947327 - f1-score (micro avg) 0.4844 +#> 2023-11-29 12:42:15,319 - 3 epochs without improvement +#> 2023-11-29 12:42:15,320 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:42:16,295 epoch 10 - iter 6/64 - loss 0.74788894 - time (sec): 0.97 - samples/sec: 196.93 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:16,969 epoch 10 - iter 12/64 - loss 0.76720373 - time (sec): 1.65 - samples/sec: 232.91 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:17,693 epoch 10 - iter 18/64 - loss 0.79161676 - time (sec): 2.37 - samples/sec: 242.76 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:18,604 epoch 10 - iter 24/64 - loss 0.78399789 - time (sec): 3.28 - samples/sec: 233.86 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:19,306 epoch 10 - iter 30/64 - loss 0.78045449 - time (sec): 3.99 - samples/sec: 240.83 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:19,999 epoch 10 - iter 36/64 - loss 0.78069342 - time (sec): 4.68 - samples/sec: 246.20 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:20,746 epoch 10 - iter 42/64 - loss 0.76349049 - time (sec): 5.43 - samples/sec: 247.69 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:21,656 epoch 10 - iter 48/64 - loss 0.76165402 - time (sec): 6.34 - samples/sec: 242.41 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:22,368 epoch 10 - iter 54/64 - loss 0.77187297 - time (sec): 7.05 - samples/sec: 245.20 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:23,081 epoch 10 - iter 60/64 - loss 0.77289468 - time (sec): 7.76 - samples/sec: 247.40 - lr: 0.100000 - momentum: 0.000000 +#> 2023-11-29 12:42:23,559 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:42:23,560 EPOCH 10 done: loss 0.7810 - lr: 0.100000 +#> 2023-11-29 12:42:24,546 DEV : loss 0.7909061908721924 - f1-score (micro avg) 0.4889 +#> 2023-11-29 12:42:25,158 - 4 epochs without improvement (above 'patience')-> annealing learning_rate to [0.05] +#> 2023-11-29 12:42:25,477 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:42:25,477 Loading model from best epoch ... +#> 2023-11-29 12:42:26,782 #> Results: #> - F-score (micro) 0.592 #> - F-score (macro) 0.5066 @@ -1559,7 +1559,7 @@

Training the Model#> macro avg 0.7200 0.5793 0.5066 250 #> weighted avg 0.7149 0.5920 0.5132 250 #> -#> 2023-11-29 09:47:10,227 ---------------------------------------------------------------------------------------------------- +#> 2023-11-29 12:42:26,782 ---------------------------------------------------------------------------------------------------- #> $test_score #> [1] 0.592

@@ -1634,18 +1634,18 @@

Get the IMDB Corpus# load the IMDB file and downsize it to 0.1 IMDB <- flair_datasets()$IMDB corpus <- IMDB()$downsample(0.1) -#> 2023-11-29 09:47:10,640 Reading data from /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced -#> 2023-11-29 09:47:10,640 Train: /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced/train.txt -#> 2023-11-29 09:47:10,640 Dev: None -#> 2023-11-29 09:47:10,640 Test: None -#> 2023-11-29 09:47:11,239 No test split found. Using 0% (i.e. 5000 samples) of the train split as test data -#> 2023-11-29 09:47:11,254 No dev split found. Using 0% (i.e. 4500 samples) of the train split as dev data -#> 2023-11-29 09:47:11,254 Initialized corpus /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced (label type name is 'sentiment') +#> 2023-11-29 12:42:27,196 Reading data from /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced +#> 2023-11-29 12:42:27,196 Train: /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced/train.txt +#> 2023-11-29 12:42:27,196 Dev: None +#> 2023-11-29 12:42:27,196 Test: None +#> 2023-11-29 12:42:27,784 No test split found. Using 0% (i.e. 5000 samples) of the train split as test data +#> 2023-11-29 12:42:27,799 No dev split found. Using 0% (i.e. 4500 samples) of the train split as dev data +#> 2023-11-29 12:42:27,799 Initialized corpus /Users/yenchiehliao/.flair/datasets/imdb_v4-rebalanced (label type name is 'sentiment') # create the label dictionary lbl_type <- 'sentiment' label_dict <- corpus$make_label_dictionary(label_type=lbl_type) -#> 2023-11-29 09:47:11,272 Computing label dictionary. Progress: -#> 2023-11-29 09:47:19,010 Dictionary created for label 'sentiment' with 2 values: POSITIVE (seen 2056 times), NEGATIVE (seen 1994 times) +#> 2023-11-29 12:42:27,816 Computing label dictionary. Progress: +#> 2023-11-29 12:42:35,340 Dictionary created for label 'sentiment' with 2 values: POSITIVE (seen 2056 times), NEGATIVE (seen 1994 times)
diff --git a/index.html b/index.html index 57eb41ad..df13349c 100644 --- a/index.html +++ b/index.html @@ -165,8 +165,9 @@

R-CMD-Check coverage codecov CodeFactor

-

{flaiR} is an R wrapper for the {flairNLP/flair} library in Python, designed specifically for R users, especially those in the social sciences. It provides easy access to the main functionalities of {flairNLP}. Developed by Developed by Zalando Research in Berlin, Flair NLP offers intuitive interfaces and exceptional multilingual support, particularly for various embedding frameworks, transformers and state-of-the-art natural language processing tasks to analyze your text, such as named entity recognition, sentiment analysis, part-of-speech tagging, biomedical data, sense disambiguation, and classification, with support for a rapidly growing number of languages in the community. For a comprehensive understanding of the {flairNLP/flair} architecture, you can refer to the research article ‘Contextual String Embeddings for Sequence Labeling’ and the official manual written for its Python implementation.

-
+

+ {flaiR} is an R wrapper for the {flairNLP/flair} Python library, specifically tailored for R users, particularly in political science and the social sciences. flaiR provides easy access to the main functionalities of {Flair NLP}. Developed by Developed by Zalando Research in Berlin, flair NLP is a straightforward framework for state-of-the-art Natural Language Processing (NLP) and is compatible with Hugging Face. Flair offers intuitive interfaces and exceptional multilingual support, particularly for various embedding models, transformers and state-of-the-art NLP tasks to analyze texts, such as named entity recognition, sentiment analysis, part-of-speech tagging, with support for a rapidly growing number of language models in the community.

+

For a comprehensive understanding of the {flairNLP/flair} architecture and NLP tagging models by Zalando Research, you can refer to the research article ‘Contextual String Embeddings for Sequence Labeling’ and the official manual written for its Python implementation. The community support for more languages is rapidly expanding. This unofficial platform provides R users with documentation, examples, and tutorials for using Flair NLP. The goal is to make it easier for R users to access the powerful NLP tools provided by Flair NLP.


@@ -183,7 +184,7 @@

Installation via GitHub

  • Anaconda or miniconda (highly recommended)

    • We have tested flaiR using CI/CD with GitHub Actions, conducting integration tests across various operating syste These tests include intergration between R versions 4.2.1, 4.3.2, and 4.2.0 and Python 3.10.x. The testing also covers environments with flair NLP and PyTorch (given that Flair NLP is built on Torch). For stable usage, we strongly recommend installing these specific versions.

    -

    When first installed, {flaiR} automatically detects whether you have Python 3.8 or higher. If not, it will skip the automatic installation of Python and flair NLP. In this case, you will need to manually install it yourself and reload {flaiR} again. If you have correct Python installed, the {flaiR} will automatically install flair Python NLP in your global environment. If you are using {reticulate}, {flaiR} will typically assume the r-reticulate environment by default. At the same time, you can use py_config() to check the location of your environment. Please note that flaiR will directly install flair NLP in the Python environment that your R is using. This environment can be adjusted through RStudio by navigating to Tools -> Global Options -> Python. If there are any issues with the installation, feel free to ask in the Discussion .

    +

    When first installed, {flaiR} automatically detects whether you have Python 3.8 or higher. If not, it will skip the automatic installation of Python and flair NLP. In this case, you will need to manually install it yourself and reload {flaiR} again. If you have correct Python installed, the {flaiR} will automatically install flair Python NLP in your global environment. If you are using {reticulate}, {flaiR} will typically assume the r-reticulate environment by default. At the same time, you can use py_config() to check the location of your environment. Please note that flaiR will directly install flair NLP in the Python environment that your R is using. This environment can be adjusted through RStudio by navigating to Tools -> Global Options -> Python. If there are any issues with the installation, feel free to ask in the Discussion .

    First, understanding which Python environment your RStudio is using is very important. We advise you to confirm which Python environment RStudio is using. You can do this by checking with reticulate::py_config() or manually via Tools -> Global Options -> Python.

     install.packages("reticulate")
@@ -250,15 +251,15 @@ 
    
 
    Step 2 Preprocess Data and Corpus Object
    
 
     corpus <- Corpus(train=train, test=test)
-#> 2023-11-29 09:32:41,983 No dev split found. Using 0% (i.e. 282 samples) of the train split as dev data
    
+#> 2023-11-29 12:28:25,704 No dev split found. Using 0% (i.e. 282 samples) of the train split as dev data

    Step 3 Create Classifier Using Transformer

     document_embeddings <- TransformerDocumentEmbeddings('distilbert-base-uncased', fine_tune=TRUE)

    First, $make_label_dictionary function is used to automatically create a label dictionary for the classification task. The label dictionary is a mapping from label to index, which is used to map the labels to a tensor of label indices. expcept classifcation task, flair also supports other label types for training custom model, such as ner, pos and sentiment.

     label_dict <- corpus$make_label_dictionary(label_type="classification")
-#> 2023-11-29 09:32:43,454 Computing label dictionary. Progress:
-#> 2023-11-29 09:32:43,504 Dictionary created for label 'classification' with 2 values: 0 (seen 1321 times), 1 (seen 1213 times)
    +#> 2023-11-29 12:28:27,310 Computing label dictionary. Progress: +#> 2023-11-29 12:28:27,361 Dictionary created for label 'classification' with 2 values: 0 (seen 1322 times), 1 (seen 1212 times)

    Alternatively, you can also create a label dictionary manually. The following code creates a label dictionary with two labels, 0 and 1, and maps them to the indices 0 and 1 respectively.

     # load Dictionary object from flair_data
@@ -409,14 +410,14 @@ 
    
 
     classifier$predict(sentence)
 print(sentence)
-#> Sentence[55]: "Ladies and gentlemen, I stand before you today not just as a legislator, but as a defender of our very way of life! We are facing a crisis of monumental proportions, and if we don't act now, the very fabric of our society will unravel before our eyes!" → 1 (0.6431)
    
+#> Sentence[55]: "Ladies and gentlemen, I stand before you today not just as a legislator, but as a defender of our very way of life! We are facing a crisis of monumental proportions, and if we don't act now, the very fabric of our society will unravel before our eyes!" → 1 (0.5151)

    sentence$labels is a list of labels, each of which has a value and a score. The value is the label itself, and the score is the probability of the label. The label with the highest score is the predicted label.

     sentence$labels[[1]]$value
 #> [1] "1"
     sentence$labels[[1]]$score
-#> [1] 0.6430542
    +#> [1] 0.5150982

    Step 7 Reload the Model with the Best Performance

    When you train the model with save_final_model=TRUE, the model with the best performance on the development set will be saved in the output directory. You can reload the model with the best performance using the load function.

    @@ -471,7 +472,6 @@ 
    
 #> 5               Blaming Political Opponents                1 0.999097466468811


    -

    Secondly, to facilitate more efficient use for social science research, {flairR} expands {flairNLP/flair}’s core functionality for working with three major functions to extract features in a tidy and fast format– data.table in R.

    @@ -479,7 +479,7 @@

    Flair NLP also provides a set of functions to perform NLP tasks, such as named entity recognition, sentiment analysis, and part-of-speech tagging.

    -

    First, we load the data and the model and perform NER task on the text below.

    +

    First, we load the data and the model to perform NER task on the text below.

    Yesterday, Dr. Jane Smith spoke at the United Nations in New York. She discussed climate change and its impact on global economies. The event was attended by representatives from various countries including France and Japan. Dr. Smith mentioned that by 2050, the world could see a rise in sea level by approximately 2 feet. The World Health Organization (WHO) has pledged $50 million to combat the health effects of global warming. In an interview with The New York Times, Dr. Smith emphasized the urgent need for action. Later that day, she flew back to London, arriving at 10:00 PM GMT.

    @@ -489,7 +489,7 @@

    # load the model flair NLP already trained for us tagger <- Classifier$load('ner') -#> 2023-11-29 09:32:46,364 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP> +#> 2023-11-29 12:28:30,624 SequenceTagger predicts: Dictionary with 20 tags: <unk>, O, S-ORG, S-MISC, B-PER, E-PER, S-LOC, B-ORG, E-ORG, I-PER, S-PER, B-MISC, I-MISC, E-MISC, I-ORG, B-LOC, E-LOC, I-LOC, <START>, <STOP> # make a sentence object text <- "Yesterday, Dr. Jane Smith spoke at the United Nations in New York. She discussed climate change and its impact on global economies. The event was attended by representatives from various countries including France and Japan. Dr. Smith mentioned that by 2050, the world could see a rise in sea level by approximately 2 feet. The World Health Organization (WHO) has pledged $50 million to combat the health effects of global warming. In an interview with The New York Times, Dr. Smith emphasized the urgent need for action. Later that day, she flew back to London, arriving at 10:00 PM GMT." @@ -501,7 +501,8 @@

    # print sentence with predicted tags print(sentence) #> Sentence[115]: "Yesterday, Dr. Jane Smith spoke at the United Nations in New York. She discussed climate change and its impact on global economies. The event was attended by representatives from various countries including France and Japan. Dr. Smith mentioned that by 2050, the world could see a rise in sea level by approximately 2 feet. The World Health Organization (WHO) has pledged $50 million to combat the health effects of global warming. In an interview with The New York Times, Dr. Smith emphasized the urgent need for action. Later that day, she flew back to London, arriving at 10:00 PM GMT." → ["Jane Smith"/PER, "United Nations"/ORG, "New York"/LOC, "France"/LOC, "Japan"/LOC, "Smith"/PER, "World Health Organization"/ORG, "WHO"/ORG, "The New York Times"/ORG, "Smith"/PER, "London"/LOC, "GMT"/MISC]

    -

    Alternatively, the expanded features in flaiR can be used to perform and extract features from the sentence object in a tidy format.

    +

    Alternatively, to facilitate more efficient use for social science research, {flairR} expands {flairNLP/flair}’s core functionality for working with three major functions to extract features in a tidy and fast format– data.table in R.

    +

    The expanded features in flaiR can be used to perform and extract features from the sentence object in a tidy format.