remove unecessary file

.ipynb_checkpoints/app-checkpoint.py

@@ -1,109 +0,0 @@
-import streamlit as st
-from transformers import AutoTokenizer, AutoModelForCausalLM, set_seed
-from transformers import pipeline
-import torch
-import json
-import pandas as pd
-
-@st.cache(allow_output_mutation=True)
-def load_tokenizer(model_ckpt):
-    return AutoTokenizer.from_pretrained(model_ckpt)
-
-@st.cache(allow_output_mutation=True)
-def load_model(model_ckpt):
-    model = AutoModelForCausalLM.from_pretrained(model_ckpt, low_cpu_mem_usage=True)
-    return model
-
-@st.cache()
-def load_examples():
-    with open("examples.json", "r") as f:
-        examples = json.load(f)
-    return examples
-
-st.set_page_config(page_icon=':laptop:', layout="wide")
-           
-
-st.sidebar.header("Models")
-models = ["CodeParrot", "OPT", "InCoder"]
-selected_models = st.sidebar.multiselect('Select code generation models to compare:', 
-                         models,
-                         default=["CodeParrot"])
-st.sidebar.header("Tasks")
-tasks = [" ", "Model evaluation", "Pretraining datasets", "Model architecture", "Code generation"]
-selected_task = st.sidebar.selectbox("Select a task:", tasks)
-
-
-tokenizer1 = load_tokenizer("lvwerra/codeparrot")
-model1 = load_model("lvwerra/codeparrot")
-tokenizer2 = load_tokenizer("facebook/incoder-1B")
-model2 = load_model("facebook/incoder-1B")
-#tokenizer3 = load_tokenizer("facebook/opt-1.3b")
-#model3 = load_model("facebook/opt-1.3b")
-pipelines = {}
-for element in models:
-    if element == "CodeParrot":
-        pipelines[element] = pipeline("text-generation", model=model1, tokenizer=tokenizer1)
-    elif element == "InCoder":
-        tokenizer = load_tokenizer("facebook/incoder-1B")
-        model = load_model("facebook/incoder-1B")
-        pipelines[element] = pipeline("text-generation", model=model2, tokenizer=tokenizer2)
-    #else:
-    #    tokenizer = load_tokenizer("facebook/opt-1.3b")
-    #    model = load_model("facebook/opt-1.3b")
-    #    pipelines[element] = pipeline("text-generation", model=model3, tokenizer=tokenizer3)
-        
-examples = load_examples()
-example_names = [example["name"] for example in examples]
-name2id = dict([(name, i) for i, name in enumerate(example_names)])
-set_seed(42)
-gen_kwargs = {}
-
-if selected_task == " ":
-    st.title("Code Generation Models comparison")
-    with open("intro.txt", "r") as f:
-        intro = f.read()
-    st.markdown(intro)
-elif selected_task == "Pretraining datasets":
-    st.title("Pretraining datasets 📚")
-    st.markdown("Preview of some code files from Github repositories")   
-    df = pd.read_csv("preview-github-data.csv")
-    st.dataframe(df)
-    for model in selected_models:
-        with open(f"datasets/{model.lower()}.txt", "r") as f:
-            text = f.read()
-        st.markdown(f"### {model}:")
-        st.markdown(text)   
-elif selected_task == "Model architecture":
-    st.title("Model architecture 🔨")
-    for model in selected_models:
-        with open(f"architectures/{model.lower()}.txt", "r") as f:
-            text = f.read()
-        st.markdown(f"## {model}:")
-        st.markdown(text) 
-elif selected_task == "Model evaluation":
-    st.title("Code models evaluation 📊")
-    with open("evaluation/intro.txt", "r") as f:
-        intro = f.read()
-    st.markdown(intro)
-elif selected_task == "Code generation":
-    st.title("Code generation 💻")
-    st.sidebar.header("Examples")
-    selected_example = st.sidebar.selectbox("Select one of the following examples:", example_names)
-    example_text = examples[name2id[selected_example]]["value"]
-    default_length = examples[name2id[selected_example]]["length"]
-    st.sidebar.header("Generation settings")
-    gen_kwargs["do_sample"] = st.sidebar.radio("Decoding strategy:",  ["Greedy", "Sample"]) == "Sample"
-    gen_kwargs["max_new_tokens"] = st.sidebar.slider("Number of tokens to generate:", value=default_length, min_value=8, step=8, max_value=256)
-    if gen_kwargs["do_sample"]:
-        gen_kwargs["temperature"] = 0.2
-        gen_kwargs["top_k"] = 0
-        gen_kwargs["top_p"] = 0.95
-    gen_prompt = st.text_area("Generate code with prompt:", value=example_text, height=220,).strip()
-    if st.button("Generate code!"):
-        with st.spinner("Generating code..."):
-            for model in selected_models:
-                if model != "OPT":
-                    pipe = pipelines[model]
-                    generated_text = pipe(gen_prompt, **gen_kwargs)[0]['generated_text']
-                    st.markdown(f"{model}:")
-                    st.code(generated_text)