add models

app.py

@@ -10,9 +10,9 @@ model_name = "microsoft/xlm-align-base"
 tokenizer_name = "microsoft/xlm-align-base"
 num_labels = 78
 num_intents = 23
-checkpoint_path = "/content/tamil_atis/xlm_align_base.bin"
-intent_encoder_path = "/content/tamil_atis/intent_annotations.npy"
-ner_encoder_path = "/content/tamil_atis/ner_annotations.npy"
+checkpoint_path = "tamilatis/models/xlm_align_base.bin"
+intent_encoder_path = "tamilatis/models/intent_classes.npy"
+ner_encoder_path = "tamilatis/models/ner_classes.npy"
 
 def predict_function(text):
   label_encoder = LabelEncoder()

models/xlm_align_base.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bc9b69ea50334b1699e06a5ed1afd476b8f8d132d2165ea7bf38fd0a26181b4
+size 1110211757

requirements.txt

@@ -0,0 +1,13 @@
+accelerate==0.10.0
+gradio==3.0.20
+huggingface_hub==0.8.1
+hydra-core==1.2.0
+numpy==1.21.6
+omegaconf==2.2.2
+pandas==1.3.5
+scikit_learn==0.24.1
+seqeval==1.2.2
+torch==1.11.0+cu113
+torchmetrics==0.9.1
+tqdm==4.64.0
+transformers==4.20.1

tamilatis/main.py

@@ -2,7 +2,8 @@ import logging
 import os
 import pickle
 
-import wandb
+#import wandb
+import numpy as np
 import hydra
 import pandas as pd
 import torch.nn as nn
@@ -68,9 +69,11 @@ def main(cfg):
     # convert string labels to int
     label_encoder = LabelEncoder()
     label_encoder.fit(annotations)
+    np.save("/content/tamilatis/models/tamilatis/ner_classes.npy",label_encoder.classes_)
 
     intent_encoder = LabelEncoder()
     intent_encoder.fit(intents)
+    np.save("/content/tamilatis/models/tamilatis/intent_classes.npy",intent_encoder.classes_)
 
     train_ds = ATISDataset(
         train_data, cfg.model.tokenizer_name, label_encoder, intent_encoder