updated app

app.py

@@ -23,13 +23,13 @@ def load_model(model_name):
         model.eval()
     return tokenizer,model
 
-@st.cache
+@st.cache(show_spinner=False)
 def load_data(sentence_num):
     df = pd.read_csv('tsne_out.csv')
     df = df.loc[lambda d: (d['sentence_num']==sentence_num)&(d['iter_num']<1000)]
     return df
 
-@st.cache
+@st.cache(show_spinner=False)
 def mask_prob(model,mask_id,sentences,position,temp=1):
     masked_sentences = sentences.clone()
     masked_sentences[:, position] = mask_id
@@ -37,7 +37,7 @@ def mask_prob(model,mask_id,sentences,position,temp=1):
         logits = model(masked_sentences)[0]
     return F.log_softmax(logits[:, position] / temp, dim = -1)
 
-@st.cache
+@st.cache(show_spinner=False)
 def sample_words(probs,pos,sentences):
     candidates = [[tokenizer.decode([candidate]),torch.exp(probs)[0,candidate].item()]
                   for candidate in torch.argsort(probs[0],descending=True)[:10]]
@@ -53,7 +53,7 @@ def run_chains(tokenizer,model,mask_id,input_text,num_steps):
     sentence = init_sent.clone()
     data_list = []
     st.sidebar.write('Generating samples...')
-    st.sidebar.write('This takes ~30 seconds for 1000 steps with ~10 token sentences')
+    st.sidebar.write('This takes ~1 min for 1000 steps with ~10 token sentences')
     chain_progress = st.sidebar.progress(0)
     for step_id in range(num_steps):
         chain_progress.progress((step_id+1)/num_steps)
@@ -66,6 +66,7 @@ def run_chains(tokenizer,model,mask_id,input_text,num_steps):
 @st.cache(suppress_st_warning=True,show_spinner=False)
 def run_tsne(chain):
     st.sidebar.write('Running t-SNE...')
+    st.sidebar.write('This takes ~1 min for 1000 steps with ~10 token sentences')
     chain = chain.assign(cleaned_sentence=chain.sentence.str.replace(r'\[CLS\] ', '',regex=True).str.replace(r' \[SEP\]', '',regex=True))
     sentence_model = load_sentence_model()
     sentence_embeddings = sentence_model.encode(chain.cleaned_sentence.to_list(), show_progress_bar=False)