Skip to content

HAL quality checks - Part 2 #130

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
imenelydiaker wants to merge 6 commits into
base: main
Choose a base branch
from
Open

HAL quality checks - Part 2 #130

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
902854f
update LDA script
imenelydiaker Jun 5, 2024
0e1ba7e
update lda script
imenelydiaker Jun 5, 2024
7760d63
update scripts
imenelydiaker Jun 5, 2024
a917da7
add baselines results
imenelydiaker Jun 5, 2024
446f029
update script and scores for bert
imenelydiaker Jun 6, 2024
4c9397b
update files
imenelydiaker Jun 6, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,4 +4,4 @@ __pycache__
ChromaDB
models
.vscode
hal_baseline*/
hal_baseline_*/
63 changes: 63 additions & 0 deletions script_mteb_french/quality_checks/hal/hal_baseline/scores.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,63 @@
[
{
"score": {
"eval_loss": 1.0339970588684082,
"eval_f1": 0.6143314762314238,
"eval_runtime": 54.7352,
"eval_samples_per_second": 143.783,
"eval_steps_per_second": 8.989,
"epoch": 5.0
},
"seed": 42
},{
"score": {
"eval_loss": 1.0847108364105225,
"eval_f1": 0.6004042539017442,
"eval_runtime": 54.9377,
"eval_samples_per_second": 143.253,
"eval_steps_per_second": 8.956,
"epoch": 5.0
},
"seed": 2024
},{
"score": {
"eval_loss": 1.0931035280227661,
"eval_f1": 0.5940237137849953,
"eval_runtime": 54.5868,
"eval_samples_per_second": 144.174,
"eval_steps_per_second": 9.013,
"epoch": 5.0
},
"seed": 5050
}
]{
"score": {
"eval_loss": 1.083167552947998,
"eval_f1": 0.609011590210188,
"eval_runtime": 54.7637,
"eval_samples_per_second": 143.708,
"eval_steps_per_second": 8.984,
"epoch": 5.0
},
"seed": 42
}{
"score": {
"eval_loss": 1.0764442682266235,
"eval_f1": 0.6106352447263774,
"eval_runtime": 54.7671,
"eval_samples_per_second": 143.699,
"eval_steps_per_second": 8.983,
"epoch": 5.0
},
"seed": 42
}{
"score": {
"eval_loss": 1.08128821849823,
"eval_f1": 0.6074891108370322,
"eval_runtime": 54.5086,
"eval_samples_per_second": 144.381,
"eval_steps_per_second": 9.026,
"epoch": 5.0
},
"seed": 42
}
4 changes: 2 additions & 2 deletions script_mteb_french/quality_checks/hal/hal_baseline_bert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,10 +81,10 @@ def compute_metrics(eval_pred):
if not os.path.exists(args.model_dir):
os.makedirs(args.model_dir)

with open(f"{args.output_dir}/scores.json", "w") as f:
with open(f"{args.output_dir}/scores.json", "a+") as f:
json.dump({
"score": score,
"seed": args.model_seed,
"seed": args.dataset_seed,
}, f, indent=2)

trainer.save_model(f"{args.model_dir}_{args.dataset_seed}")
Expand Down
28 changes: 19 additions & 9 deletions script_mteb_french/quality_checks/hal/hal_baseline_lda.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,8 @@
from datasets import load_dataset

import spacy
import nltk
import spacy.cli
nltk.download('stopwords')
from nltk.corpus import stopwords

Expand All @@ -10,24 +12,32 @@

DATASET = "lyon-nlp/clustering-hal-s2s"
SEED = 42
STOPWORDS = stopwords.words("french") + stopwords.words("english")
STOPWORDS = stopwords.words("french") # + stopwords.words("english")

try:
nlp = spacy.load('fr_core_news_sm', disable = ['parser','ner'])
except FileNotFoundError:
spacy.cli.download("fr_core_news_sm")
nlp = spacy.load('fr_core_news_md', disable = ['parser','ner'])


dataset = load_dataset(DATASET, name="mteb_eval", split="test")
dataset = dataset.class_encode_column("domain")
num_classes = dataset.features["domain"].num_classes
id2label = {k: dataset.features["domain"].int2str(k) for k in range(num_classes)}
dataset = dataset.train_test_split(test_size=0.3, shuffle=True, stratify_by_column="domain", seed=SEED)

X_train, y_train = dataset["train"]["title"], dataset["train"]["domain"]
X_test, y_test = dataset["test"]["title"], dataset["test"]["domain"]
texts, domains = dataset["title"], dataset["domain"]

docs = nlp.pipe(texts)

def tokenize_text(doc):
return [token.lemma_.lower() for token in doc if token not in STOPWORDS]

tokenized_X_train = [text.split() for text in X_train]
tokenized_X_test = [text.split() for text in X_test]
tokenized_texts = [tokenize_text(doc) for doc in docs]
print(tokenized_texts[:5])

common_dictionary = Dictionary(tokenized_X_train)
common_corpus = [common_dictionary.doc2bow(text) for text in tokenized_X_train]
common_dictionary = Dictionary(tokenized_texts)
common_corpus = [common_dictionary.doc2bow(text) for text in tokenized_texts]

lda = LdaModel(
common_corpus,
Expand All @@ -38,6 +48,6 @@
print(f"Perplexity: {lda.log_perplexity(common_corpus)}")

coherence_lda = CoherenceModel(
model=lda, texts=tokenized_X_train,
model=lda, texts=tokenized_texts,
dictionary=common_dictionary, coherence='c_v')
print(f"Coherence: {coherence_lda.get_coherence()}")
7 changes: 1 addition & 6 deletions script_mteb_french/quality_checks/hal/hal_baseline_tfidf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,16 +30,11 @@ def main(args):
X_train, y_train = dataset["train"]["title"], dataset["train"]["domain"]
X_test, y_test = dataset["test"]["title"], dataset["test"]["domain"]

estimators = [LogisticRegression(max_iter=1000), RandomForestClassifier(), SVC()]
estimators = [LogisticRegression(max_iter=1000), SVC()]
parameters = [
{
"estimator__C": [0.1, 1, 10],
},
{
"estimator__n_estimators": [200, 300],
"estimator__bootstrap": [True, False],
"estimator__class_weight": ["balanced", "balanced_subsample"],
},
{
"estimator__C": [0.1, 1, 10],
"estimator__kernel": ["linear", "poly", "rbf", "sigmoid"],
Expand Down
3 changes: 3 additions & 0 deletions script_mteb_french/quality_checks/hal/requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,3 @@
langdetect>=1.0.9
gensim>=4.3.2
spacy>=3.7.5
6 changes: 1 addition & 5 deletions script_mteb_french/quality_checks/hal/run_bert.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,11 +4,7 @@ seeds=(
5050
)

cd script_mteb_french/quality_checks/hal

for seed in "${seeds[@]}";
do
python hal_baseline_bert.py --dataset_seed $seed
python hal_baseline_bert.py --dataset_seed $seed --epochs 5 --batch_size 32 --lr 1e-4
done

cd ../../..
1 change: 1 addition & 0 deletions script_mteb_french/quality_checks/hal/run_lda.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
python hal_baseline_lda.py
4 changes: 0 additions & 4 deletions script_mteb_french/quality_checks/hal/run_tfidf.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,11 +4,7 @@ seeds=(
5050
)

cd script_mteb_french/quality_checks/hal

for seed in "${seeds[@]}";
do
python hal_baseline_tfidf.py --dataset_seed $seed
done

cd ../../..
4 changes: 4 additions & 0 deletions script_mteb_french/quality_checks/hal/tfidf_baseline/params_LogisticRegression_2024.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,4 @@
{
"estimator__C": 10,
"tfidf__max_features": 10000
}
2 changes: 1 addition & 1 deletion script_mteb_french/quality_checks/hal/tfidf_baseline/params_LogisticRegression_42.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
{
"estimator__C": 10,
"estimator__C": 1,
"tfidf__max_features": 10000
}
4 changes: 4 additions & 0 deletions script_mteb_french/quality_checks/hal/tfidf_baseline/params_LogisticRegression_5050.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,4 @@
{
"estimator__C": 10,
"tfidf__max_features": 10000
}
6 changes: 0 additions & 6 deletions script_mteb_french/quality_checks/hal/tfidf_baseline/params_RandomForestClassifier_42.json
View file
Open in desktop

This file was deleted.

6 changes: 6 additions & 0 deletions script_mteb_french/quality_checks/hal/tfidf_baseline/params_SVC_2024.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,6 @@
{
"estimator__C": 1,
"estimator__class_weight": null,
"estimator__kernel": "linear",
"tfidf__max_features": 10000
}
6 changes: 6 additions & 0 deletions script_mteb_french/quality_checks/hal/tfidf_baseline/params_SVC_42.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,6 @@
{
"estimator__C": 1,
"estimator__class_weight": null,
"estimator__kernel": "linear",
"tfidf__max_features": 10000
}
6 changes: 6 additions & 0 deletions script_mteb_french/quality_checks/hal/tfidf_baseline/params_SVC_5050.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,6 @@
{
"estimator__C": 1,
"estimator__class_weight": "balanced",
"estimator__kernel": "rbf",
"tfidf__max_features": 10000
}
75 changes: 75 additions & 0 deletions script_mteb_french/quality_checks/hal/tfidf_baseline/report_LogisticRegression_2024.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,75 @@
{
"chim": {
"precision": 0.54375,
"recall": 0.39545454545454545,
"f1-score": 0.45789473684210524,
"support": 220.0
},
"info": {
"precision": 0.5698427382053654,
"recall": 0.6292134831460674,
"f1-score": 0.5980582524271845,
"support": 979.0
},
"math": {
"precision": 0.5304878048780488,
"recall": 0.3522267206477733,
"f1-score": 0.4233576642335766,
"support": 247.0
},
"phys": {
"precision": 0.4656188605108055,
"recall": 0.4100346020761246,
"f1-score": 0.43606255749770007,
"support": 578.0
},
"scco": {
"precision": 0.3493975903614458,
"recall": 0.15591397849462366,
"f1-score": 0.21561338289962825,
"support": 186.0
},
"sde": {
"precision": 0.5275779376498801,
"recall": 0.5326876513317191,
"f1-score": 0.5301204819277109,
"support": 826.0
},
"sdu": {
"precision": 0.5843621399176955,
"recall": 0.40804597701149425,
"f1-score": 0.4805414551607445,
"support": 348.0
},
"sdv": {
"precision": 0.6847672778561354,
"recall": 0.6738376127689105,
"f1-score": 0.6792584819867087,
"support": 1441.0
},
"shs": {
"precision": 0.7087294727744166,
"recall": 0.8159203980099502,
"f1-score": 0.7585568917668826,
"support": 2010.0
},
"spi": {
"precision": 0.5385338345864662,
"recall": 0.553623188405797,
"f1-score": 0.5459742734635541,
"support": 1035.0
},
"accuracy": 0.6127064803049556,
"macro avg": {
"precision": 0.5503067656740259,
"recall": 0.49269581573470056,
"f1-score": 0.5125438178205795,
"support": 7870.0
},
"weighted avg": {
"precision": 0.6036166441353017,
"recall": 0.6127064803049556,
"f1-score": 0.6044035075282668,
"support": 7870.0
}
}
Loading