Update app.py

app.py

@@ -11,6 +11,7 @@ from transformers import (AutoModelForCausalLM, AutoModelForQuestionAnswering,
                           pipeline, top_k_top_p_filtering)
 
 
+
 st.set_page_config(page_title="Gadsby")
 st.title("Gadsby - Constrained Text Generation with Transformers")
 st.caption("By Allen Roush")
@@ -25,7 +26,7 @@ form.header("Main Settings")
 
 model_name = form.text_area("Enter the name of the pre-trained model from transformers that we are using for Text Generation", value = "gpt2")
 form.caption("This will download a new model, so it may take awhile or even break if the model is too large")
-mode = form.selectbox("What kind of constrained generation are we doing?", ["lipogram", "reverse_lipogram", "e-prime", "rhopalism", "length_constrained", "greater_than_length"])
+mode = form.selectbox("What kind of constrained generation are we doing?", ["lipogram", "reverse_lipogram", "e-prime", "rhopalism", "length_constrained", "greater_than_length", "Pangram", "rhopalism-lipogram"])
 form.caption("Lipograms mean that a letter (or substring) is not allowed in the generated string, reverse lipograms force a letter to be in the generated string")
 
 if mode == "lipogram":
@@ -42,12 +43,17 @@ elif mode == "reverse_lipogram":
 elif mode == "rhopalism":
     length_constraint = form.number_input("Enter the length that the Rhopalism shoud start with", value = 1)
     st.caption("Rhopalisms are usually reliable but sometimes you need to try generating two or three times for a perfect one")
+elif mode == "rhopalism-lipogram":
+    naughty_strings_list = st.text_area("Enter the list of strings that you don't want in each word seperated by a space", value = "E e")
+    naughty_strings = naughty_strings_list.split(" ")
+    length_constraint = form.number_input("Enter the length that the Rhopalism shoud start with", value = 1)
+    st.caption("Rhopalisms are usually reliable but sometimes you need to try generating two or three times for a perfect one")
 else:
     length_constraint = form.number_input("Enter the length should each word be restricted to (or greater/less than)", value = 5) + 1
 
 
 length = form.number_input("Select how long you want the generated text to be", value = 100)
-number_of_tokens_to_sample = form.number_input("Select how many tokens we want to search through when we do the filtering", value = 1000)
+number_of_tokens_to_sample = form.number_input("Select how many tokens we want to search through when we do the filtering", value = 25000)
 form.caption("Settings this to higher numbers will improve the experience but will cause generating to slow. Low numbers may cause lots of blank or failed generations")
 temperature = form.number_input("How spicy/interesting do we want our models output to be", value = 1.05, min_value = 0.0)
 form.caption("Setting this higher decreases the likelihood of high probability words and increases the likelihood of low probability (and presumably more interesting) words")
@@ -68,9 +74,13 @@ def isPalindrome(s):
     return s == s[::-1]
 
 
-if mode == "rhopalism":
+if mode == "rhopalism" or mode == "rhopalism-lipogram":
     rhopalism_len = length_constraint
-alphabet_string = string.ascii_lowercase
+
+
+
+nice_strings_pangram = list(string.ascii_lowercase)
+
 
 
 def get_next_word_without_e(input_sequence):
@@ -110,19 +120,38 @@ def get_next_word_without_e(input_sequence):
                 ##Mostly works
                 if len(resulting_string) == rhopalism_len:
                     return resulting_string
+            elif mode == "Pangram":
+                if any(c in nice_strings_pangram for c in resulting_string):
+                    return resulting_string
+            elif mode == "rhopalism-lipogram":
+                if len(resulting_string) == rhopalism_len:
+                    if all(nauty_string not in resulting_string for nauty_string in naughty_strings):
+                        return resulting_string
+
+
+                
     return " "
 
 
 
+
+j = 0
 i = length
 while i > 0:
+    current_digit = int(digits_of_pi[j])
     new_word = get_next_word_without_e(input_sequence= sequence)
     sequence = sequence + new_word
-    if mode == "rhopalism":
+    if mode == "rhopalism" or mode == "rhopalism-lipogram":
         rhopalism_len += 1
     i = i-1
+    if mode == "Pangram":
+        for character in sequence:
+            if character in nice_strings_pangram:
+                nice_strings_pangram.remove(character)
+    j += 1 
 
 st.write("GENERATED SEQUENCE: ")
 st.write(sequence)
+#st.write(nice_strings_pangram)