Addding lstm model

app_main_hf.py

@@ -15,7 +15,7 @@ import shutil
 import gc
 from transformers.utils.hub import TRANSFORMERS_CACHE
 
-torch.classes.__path__ = []
+torch.classes.__path__ = [] 
 
 
 try:

sentiment_analysis/config/stage1_models.json

@@ -43,5 +43,20 @@
         "device": "cpu",
         "load_function": "load_model",
         "predict_function": "predict"
+    },
+    "4": {
+        "name": "LSTM Custom Model",
+        "type": "lstm_uncased_custom",
+        "module_path": "hmv_cfg_base_stage1.model4",
+        "hf_location": "tachygraphy-microtrext-norm-org/LSTM-LV1-SentimentPolarities",
+        "tokenizer_class": "",
+        "model_class": "",
+        "problem_type": "multi_label_classification",
+        "base_model": "",
+        "base_model_class": "",
+        "num_labels": 3,
+        "device": "cpu",
+        "load_function": "load_model",
+        "predict_function": "predict"
     }
 }

sentiment_analysis/hmv_cfg_base_stage1/model4.py

@@ -0,0 +1,146 @@
+import os
+import sys
+
+sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), )))
+
+from imports import *
+
+import importlib.util
+import os
+import sys
+import joblib
+
+import torch
+import torch.nn as nn
+import torch.functional as F
+from transformers import DebertaV2Model, DebertaV2Tokenizer, AutoModel, AutoTokenizer
+import safetensors
+# from safetensors import load_file, save_file
+import json
+from huggingface_hub import hf_hub_download
+from safetensors.torch import save_file, safe_open
+
+
+import pickle
+import tensorflow as tf
+# from tensorflow.keras.preprocessing.sequence import pad_sequences
+# from keras.preprocessing.sequence import pad_sequences
+# from keras_preprocessing.sequence import pad_sequences
+# from tensorflow.keras.preprocessing.sequence import pad_sequences
+
+
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+CONFIG_STAGE1 = os.path.join(BASE_DIR, "..", "config", "stage1_models.json")
+
+MODEL_OPTIONS = {
+    "4": {
+        "name": "LSTM Custom Model",
+        "type": "lstm_uncased_custom",
+        "module_path": "hmv_cfg_base_stage1.model4",
+        "hf_location": "tachygraphy-microtrext-norm-org/LSTM-LV1-SentimentPolarities",
+        "tokenizer_class": "",
+        "model_class": "",
+        "problem_type": "multi_label_classification",
+        "base_model": "",
+        "base_model_class": "",
+        "num_labels": 3,
+        "device": "cpu",
+        "load_function": "load_model",
+        "predict_function": "predict"
+    }
+}
+
+
+model_key = "4"
+model_info = MODEL_OPTIONS[model_key]
+hf_location = model_info["hf_location"]
+
+
+@st.cache_resource
+def load_model():
+    repo_id = hf_location
+    print("Loading model 4")
+    model_path = hf_hub_download(repo_id=repo_id, filename="lstm.h5")
+    tokenizer_path = hf_hub_download(repo_id=repo_id, filename="tokenizer.pickle")
+
+    lstm_model = tf.keras.models.load_model(model_path)
+
+    with open(tokenizer_path, "rb") as handle:
+        tokenizer = pickle.load(handle)
+    print("Model 4 loaded")
+
+    return lstm_model, tokenizer
+
+
+def pad_sequences_custom(sequences, maxlen, dtype="int32", padding="pre", truncating="pre", value=0):
+    """
+    Pads each sequence to the same length (maxlen).
+
+    Args:
+        sequences (list of list of int): A list where each element is a sequence (list of integers).
+        maxlen (int): Maximum length of all sequences.
+        dtype (str): Data type of the output (default "int32").
+        padding (str): 'pre' or 'post'—whether to add padding before or after the sequence.
+        truncating (str): 'pre' or 'post'—whether to remove values from the beginning or end if sequence is too long.
+        value (int): The padding value.
+
+    Returns:
+        numpy.ndarray: 2D array of shape (number of sequences, maxlen)
+    """
+    # Initialize a numpy array with the pad value.
+    num_samples = len(sequences)
+    padded = np.full((num_samples, maxlen), value, dtype=dtype)
+    
+    for i, seq in enumerate(sequences):
+        if not seq:
+            continue  # skip empty sequences
+        if len(seq) > maxlen:
+            if truncating == "pre":
+                trunc = seq[-maxlen:]
+            elif truncating == "post":
+                trunc = seq[:maxlen]
+            else:
+                raise ValueError("Invalid truncating type: choose 'pre' or 'post'.")
+        else:
+            trunc = seq
+        if padding == "post":
+            padded[i, :len(trunc)] = trunc
+        elif padding == "pre":
+            padded[i, -len(trunc):] = trunc
+        else:
+            raise ValueError("Invalid padding type: choose 'pre' or 'post'.")
+    
+    return padded
+
+def predict(text, model, tokenizer, device, max_len=128):
+# def predict(text, model, tokenizer, max_len=128):
+    # Convert text to a sequence of integers
+    sequences = tokenizer.texts_to_sequences([text])
+    # Pad the sequence using our custom padding function
+    padded_sequences = pad_sequences_custom(sequences, maxlen=max_len, dtype="int32", value=0)
+    # Get the model's output (logits); assume shape is (1, num_classes)
+    logits = model.predict(padded_sequences, batch_size=1, verbose=0)[0]
+    print(logits)
+    # Convert logits to probabilities using the exponential and normalize (softmax)
+    # exp_logits = np.exp(logits)
+    # probabilities = exp_logits / np.sum(exp_logits)
+    
+    # Ensure the output is a 2D array: shape (1, 3)
+    probabilities = logits / logits.sum()
+    print(probabilities)
+    probabilities = np.atleast_2d(probabilities)
+    print(probabilities)
+    return probabilities
+
+
+# def predict(text, model, tokenizer, max_len=128):
+#     sequences = tokenizer.texts_to_sequences([text])
+#     # Use our custom pad_sequences function:
+#     padded_sequences = pad_sequences_custom(sequences, maxlen=max_len, dtype="int32", value=0)
+#     prediction = model.predict(padded_sequences, batch_size=1, verbose=0)[0]
+#     pred_class = np.argmax(prediction)
+#     sentiment_labels = ["Negative", "Neutral", "Positive"]
+#     probabilities = prediction / prediction.sum()
+#     return sentiment_labels[pred_class], pred_class, probabilities
+
+

sentiment_analysis/sentiment_analysis_main.py

@@ -195,7 +195,7 @@ if "model_changed" not in st.session_state:
     st.session_state.model_changed = False
 if "text_changed" not in st.session_state:
     st.session_state.text_changed = False
-if "processing" not in st.session_state:
+if "disabled" not in st.session_state:
     st.session_state.disabled = False
 
 
@@ -249,7 +249,9 @@ def show_sentiment_analysis():
                     "⚠️ Error: Model failed to load! Check model selection or configuration.")
                 st.stop()
 
-            model.to(device)
+            # model.to(device)
+            if hasattr(model, "to"):
+                model.to(device)
 
             # predictions = predict(user_input, model, tokenizer, device)
 
@@ -270,7 +272,7 @@ def show_sentiment_analysis():
         st.write(f"**Predicted Sentiment Scores:** {predictions_array}")
 
         # enable_ui()
-
+##
         # Display binary classification result
         st.write(f"**Predicted Sentiment:**")
         st.write(f"**NEGATIVE:** {binary_predictions[0]}, **NEUTRAL:** {binary_predictions[1]}, **POSITIVE:** {binary_predictions[2]}")