Updating to include NLI lora model

app.py

@@ -1,5 +1,5 @@
 import gradio as gr
-from transformers import pipeline, AutoTokenizer, AutoModel, BertForSequenceClassification
+from transformers import pipeline, AutoTokenizer, AutoModel, BertForSequenceClassification, AlbertForSequenceClassification, DebertaForSequenceClassification, AutoModelForSequenceClassification
 from peft.auto import AutoPeftModelForSequenceClassification
 from tensorboard.backend.event_processing import event_accumulator
 from peft import PeftModel
@@ -13,10 +13,6 @@ loraModel = AutoPeftModelForSequenceClassification.from_pretrained("Intradiction
 #tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased")
 tokenizer1 = AutoTokenizer.from_pretrained("albert-base-v2")
 tokenizer2 = AutoTokenizer.from_pretrained("microsoft/deberta-v3-xsmall")
-# base_model = AutoModel.from_pretrained("microsoft/deberta-v3-xsmall")
-# peft_model_id = "rajevan123/STS-Lora-Fine-Tuning-Capstone-Deberta-small"
-# model = PeftModel.from_pretrained(base_model, peft_model_id)
-# #merged_model = model.merge_and_unload()
 
 
 # Handle calls to DistilBERT------------------------------------------
@@ -42,32 +38,44 @@ def distilBERTUntrained_fn(text):
 
 
 # Handle calls to ALBERT---------------------------------------------
-ALbertUntrained_pipe = pipeline("text-classification", model="albert-base-v2")
-AlbertnoLORA_pipe = pipeline(model="Intradiction/NLI-Conventional-Fine-Tuning")
-#AlbertwithLORA_pipe = pipeline()
+base_model1 = AlbertForSequenceClassification.from_pretrained("Alireza1044/albert-base-v2-mnli")
+peft_model_id1 = "m4faisal/NLI-Lora-Fine-Tuning-10K-ALBERT"
+model1 = PeftModel.from_pretrained(model=base_model1, model_id=peft_model_id1)
+sa_merged_model1 = model1.merge_and_unload()
+bbu_tokenizer1 = AutoTokenizer.from_pretrained("Alireza1044/albert-base-v2-mnli")
+
+ALbertUntrained_pipe = pipeline("text-classification", model="Alireza1044/albert-base-v2-mnli")
+AlbertnoLORA_pipe = pipeline(model="m4faisal/NLI-Conventional-Fine-Tuning")
+AlbertwithLORA_pipe = pipeline("text-classification",model=sa_merged_model1, tokenizer=bbu_tokenizer1)
 
 #NLI models 
 def AlbertnoLORA_fn(text1, text2):
     return AlbertnoLORA_pipe({'text': text1, 'text_pair': text2})
 
 def AlbertwithLORA_fn(text1, text2):
-    return ("working2")
+    return AlbertwithLORA_pipe({'text': text1, 'text_pair': text2})
 
 def AlbertUntrained_fn(text1, text2):
     return ALbertUntrained_pipe({'text': text1, 'text_pair': text2})
 
 
 # Handle calls to Deberta--------------------------------------------
+# base_model2 = AutoModelForSequenceClassification.from_pretrained("microsoft/deberta-v3-xsmall", ignore_mismatched_sizes=True)
+# peft_model_id2 = "rajevan123/STS-Lora-Fine-Tuning-Capstone-Deberta-old-model-pipe-test_augmentation"
+# model2 = PeftModel.from_pretrained(model=base_model2, model_id=peft_model_id2)
+# sa_merged_model2 = model2.merge_and_unload()
+# bbu_tokenizer2 = AutoTokenizer.from_pretrained("microsoft/deberta-v3-xsmall")
+
 DebertaUntrained_pipe = pipeline("text-classification", model="microsoft/deberta-v3-xsmall")
 DebertanoLORA_pipe = pipeline("text-classification", model="rajevan123/STS-Conventional-Fine-Tuning")
-#DebertawithLORA_pipe = pipeline("text-classification",model=model, tokenizer=tokenizer2)
+# DebertawithLORA_pipe = pipeline("text-classification",model=sa_merged_model2, tokenizer=bbu_tokenizer2)
 
 #STS models
 def DebertanoLORA_fn(text1, text2):
     return DebertanoLORA_pipe({'text': text1, 'text_pair': text2})
 
 def DebertawithLORA_fn(text1, text2):
-    #return DebertawithLORA_pipe({'text': text1, 'text_pair': text2})
+    # return DebertawithLORA_pipe({'text': text1, 'text_pair': text2})
     return ("working2")
 
 def DebertaUntrained_fn(text1, text2):
@@ -94,7 +102,16 @@ def displayMetricStatsText():
     event_acc.Reload()
     accuracy_data = event_acc.Scalars('eval/accuracy')
     loss_data = event_acc.Scalars('eval/loss')
-    metrics = ''
+
+    #code to pull time data (very inaccurate)
+    # time_data = event_acc.Scalars('eval/runtime')
+    # Ttime = 0 
+    # for time in time_data:
+    #     Ttime+=time.value
+    # Ttime = str(round(Ttime/60,2))
+    # print(Ttime)
+
+    metrics = ("Active Training Time:  mins \n\n")
     for i in range(0, len(loss_data)):
         metrics = metrics + 'Epoch Number: ' + str(i) + '\n'
         metrics = metrics + 'Accuracy (%): ' + str(round(accuracy_data[i].value * 100, 3)) + '\n'
@@ -117,7 +134,16 @@ def displayMetricStatsTextTCLora():
     event_acc.Reload()
     accuracy_data = event_acc.Scalars('eval/accuracy')
     loss_data = event_acc.Scalars('eval/loss')
-    metrics = ''
+    
+    #code to pull time data (very inaccurate)
+    # time_data = event_acc.Scalars('eval/runtime')
+    # Ttime = 0 
+    # for time in time_data:
+    #     Ttime+=time.value
+    # Ttime = str(round(Ttime/60,2))
+    # print(event_acc.Tags())
+
+    metrics = ("Active Training Time:  mins \n\n")
     for i in range(0, len(loss_data)):
         metrics = metrics + 'Epoch Number: ' + str(i) + '\n'
         metrics = metrics + 'Accuracy (%): ' + str(round(accuracy_data[i].value * 100, 3)) + '\n'
@@ -127,7 +153,7 @@ def displayMetricStatsTextTCLora():
 
 def displayMetricStatsTextNLINoLora():
     #file_name = 'events.out.tfevents.NLI-Conventional.1'
-    file_name = hf_hub_download(repo_id="Intradiction/NLI-Conventional-Fine-Tuning", filename="runs/Feb15_15-44-46_d5fea890778d/events.out.tfevents.1708014218.d5fea890778d.473.1")
+    file_name = hf_hub_download(repo_id="m4faisal/NLI-Conventional-Fine-Tuning", filename="runs/Mar20_23-18-22_a7cbf6b28344/events.out.tfevents.1710976706.a7cbf6b28344.5071.0")
     event_acc = event_accumulator.EventAccumulator(file_name,
     size_guidance={
     event_accumulator.COMPRESSED_HISTOGRAMS: 500,
@@ -149,7 +175,8 @@ def displayMetricStatsTextNLINoLora():
     return metrics
 
 def displayMetricStatsTextNLILora():
-    file_name = 'events.out.tfevents.NLI-Lora.0'
+    #file_name = 'events.out.tfevents.NLI-Lora.0'
+    file_name = hf_hub_download(repo_id="m4faisal/NLI-Lora-Fine-Tuning-10K", filename="runs/Mar20_18-07-52_87caf1b1d04f/events.out.tfevents.1710958080.87caf1b1d04f.7531.0")
     event_acc = event_accumulator.EventAccumulator(file_name,
     size_guidance={
     event_accumulator.COMPRESSED_HISTOGRAMS: 500,