Get an error while trying to use API for this model.

#2
by XiaotianHe - opened

I tried input text for both "Encryption algorithm AES [MASK_ID]" and "Encryption algorithm AES [MASK]", both return following error:

"error, No mask_token (

Please advise.

error, No mask_token (

error, No mask_token (mask) found on the input

NeuLab @ LTI/CMU org

Hi @XiaotianHe ,
Thank you for your interest in our work!
The mask token is

Uri

NeuLab @ LTI/CMU org

< m a s k >

But remove the spaces 😀

Can you give me a complete sample input?

I tried input text "Encryption algorithm AES [Uri]" and still got the same error.

NeuLab @ LTI/CMU org

Here is a sample output:

while (<mask>) {

Best,
Uri

I would like to have a sample input, not output, to retrieve a complete java code from your model. Such as "Encryption algorithm AES [Uri]" or "Encryption algorithm AES

I would like to have a sample input, not output, to retrieve a complete java code for a specified algorithm from your model. Such as "Encryption algorithm AES (mask)". So I can see a complete set of java code to get AES implemented.

NeuLab @ LTI/CMU org
edited May 5, 2023

Sorry, I meant "a sample input". while (<mask>) { is a sample input.

But this model was not designed to "retrieve a complete java code" so I'm not sure what do you mean.

For retrieving code snippets and documentation, see our another project DocPrompting: https://github.com/shuyanzhou/docprompting

Best,
Uri

All I got is:
[{"score":0.43174809217453003,"token":29225,"token_str":"true","sequence":"while (true) {"},{"score":0.14779505133628845,"token":131,"token_str":";","sequence":"while (;) {"},{"score":0.10204089432954788,"token":48640,"token_str":";;","sequence":"while (;;) {"},{"score":0.011685648001730442,"token":9226,"token_str":"this","sequence":"while (this) {"},{"score":0.011480224318802357,"token":118,"token_str":"i","sequence":"while (i) {"}]

How would this help me in java coding? I thought your model is supposed to help with java programming.

NeuLab @ LTI/CMU org

No :-)

this model is designed for java code similarity.

For java programming, see https://github.com/VHellendoorn/Code-LMs which is available on huggingface as this model: https://huggingface.co/NinedayWang/PolyCoder-2.7B

Best,
Uri

Uri,

I looked into the model NinedayWang/PolyCoder-2.7B, I still could not get a useful java code. Here is my input and output:

Input: Give me a Java code for Encryption algorithm AES-256
Output: [{"generated_text":"Give me a Java code for Encryption algorithm AES-256.\n "}]

Do you know how to get some useful source codes?

Thanks
Xiaotian

NeuLab @ LTI/CMU org

That model was trained on code completion, so I expect it to work better if you gave it the beginning of the code, for example the function signature, and then your request as a javadoc.

Can you give me an example?

Thanks for sharing. I read it carefully, but still cannot find a sample input. Can you please simply share a sample input that leads to generating a useful java code?

NeuLab @ LTI/CMU org

For Polycoder? Here is a sample input:

def binarySearch(arr, left, right, x):
    mid = (left +

I got following result by running your sample input:

[{"generated_text":"def binarySearch(arr, left, right, x):\r\n mid = (left + right"}]

It only adds "right", and does not generate a complete code at all. Is that what I am supposed to see?

By running following input:
def binarySearch(arr, left, right, x):
mid = (left +

I got following output:
[{"generated_text":"def binarySearch(arr, left, right, x): mid = (left + right"}]

By running following input:
def binarySearch(arr, left, right, x):
mid = (left +

I got following output:
[{"generated_text":"def binarySearch(arr, left, right, x): mid = (left + right"}]

By running following input:
def binarySearch(arr, left, right, x):
mid = (left +

I got following output:
[{"generated_text":"def binarySearch(arr, left, right, x): mid = (left + right"}]

NeuLab @ LTI/CMU org

Hi @XiagotianHe,

I just ran the PolyCoder model using the exact code here: https://github.com/VHellendoorn/Code-LMs#october-2022---polycoder-is-available-on-huggingface and got following 4 completions:

def binarySearch(arr, left, right, x):
    mid = (left + right) // 2
    if arr[mid] == x:
        return mid
def binarySearch(arr, left, right, x):
    mid = (left + right) // 2
    if arr[mid] > x:
        return binarySearch(arr
def binarySearch(arr, left, right, x):
    mid = (left + right) // 2
    if arr[mid] < x:
        return -1
def binarySearch(arr, left, right, x):
    mid = (left + right) / 2
    if arr[mid] > x:
        return binarySearch(arr

How did you run it? Where is this json that you pasted coming from?

import org.json.JSONObject;
import org.json.JSONArray;
import org.apache.http.HttpEntity;
import org.apache.http.HttpResponse;
import org.apache.http.client.HttpClient;
import org.apache.http.client.methods.HttpPost;
import org.apache.http.client.methods.HttpGet;
import org.apache.http.entity.StringEntity;
import org.apache.http.impl.client.HttpClientBuilder;
import org.apache.http.util.EntityUtils;
import com.google.gson.Gson;
import com.google.gson.JsonArray;
import com.google.gson.JsonElement;
import com.google.gson.JsonObject;
import com.google.gson.JsonParser;
import com.google.gson.*;

public class ChatGPT_Code {

private static final String API_ENDPOINT = "https://api-inference.huggingface.co/models/";
    

private static final int max_length=500;

public static void main(String[] args) throws Exception{
                  
    
    // Generate response from ChatGPT model
    generateResponse();
    
    // Print response
    //System.out.println(response);
}

public static void generateResponse() throws Exception{
    // Create JSON request
    
    
    //{"error":"Model gpt2-xl is currently loading","estimated_time":20.0}
    //try {} catch (Exception) {...}
    //Exception in thread "main" java.net.SocketTimeoutException: Read timed out
    
    String model=null;
    model="NinedayWang/PolyCoder-0.4B";

    
    Utility u=new Utility();
    String inputText="def binarySearch(arr, left, right, x):" + 
            "    mid = (left +";

    
    JSONObject jsonRequest = new JSONObject();
    jsonRequest.put("inputs", inputText);
    System.out.println(jsonRequest.toString());
    //jsonRequest.put("max_length", max_length);
    //jsonRequest.put("parameters", new JSONObject().put("max_length", max_length));
    
    // Send request to API endpoint
    HttpClient httpClient = HttpClientBuilder.create().build();
    HttpPost request = new HttpPost(API_ENDPOINT+model);
    request.addHeader("Authorization", "Bearer Key");
    request.addHeader("Content-Type", "application/json");
    request.addHeader("X-Device-Type", "gpu");
    StringEntity params = new StringEntity(jsonRequest.toString(), "UTF-8");
    request.setEntity(params);
    HttpResponse response;
    try {
        response = httpClient.execute(request);
        HttpEntity entity = response.getEntity();
        String jsonString = EntityUtils.toString(entity);
        System.out.println(jsonString);
        
        Gson gson = new Gson();
        JsonParser parser = new JsonParser();
        JsonElement jsonElement = parser.parse(jsonString);

        if (jsonElement.isJsonArray()) {
            // JSON is an array
            JsonArray jsonArray = jsonElement.getAsJsonArray();
            for (JsonElement element : jsonArray) {
                JsonObject jsonObject = element.getAsJsonObject();
                for (String key : jsonObject.keySet()) {
                    JsonElement element1 = jsonObject.get(key);
                    String value = element1.getAsString();
                    System.out.println(key+", "+value);
                    if (key.equalsIgnoreCase("generated_text") && value.startsWith(inputText)) {
                        value=value.substring(inputText.length()+1);
                        //value=value.substring(0, value.lastIndexOf(".")+1);
                        int index1=value.lastIndexOf(".");
                        int index2=value.lastIndexOf("!");
                        int index3=value.lastIndexOf("?");
                        int index0=-1;
                        if (index0<index1) index0=index1;
                        if (index0<index2) index0=index2;
                        if (index0<index3) index0=index3;
                        value=value.substring(0, index0+1);
                        value=u.ReplaceStr1(value, "\n", "<br>");
                        System.out.println(value);
                    }
                }
            }
        } else if (jsonElement.isJsonObject()) {
            // JSON is an object
            JsonObject jsonObject = jsonElement.getAsJsonObject();
            for (String key : jsonObject.keySet()) {
                JsonElement element = jsonObject.get(key);
                String value = element.getAsString();
                System.out.println(key+", "+value);
            }
        } else {
            System.out.println("Invalid JSON");
        } 

        
    } catch (Exception e) {
        e.printStackTrace();
    }
}

}

{"inputs":"def binarySearch(arr, left, right, x): mid = (left +"}

NeuLab @ LTI/CMU org

This is not our code and I cannot support it. We just released the model, and we demonstrate that the model works. I hope you understand

NeuLab @ LTI/CMU org

This is not our code and we cannot debug it.
We just released the model and demonstrated that the model works. I hope you understand.

How about the input json? Does it look correct?
{"inputs":"def binarySearch(arr, left, right, x): mid = (left +"}

NeuLab @ LTI/CMU org

If it wouldn't have looked correct to me I wouldn't have provided it to you :-)
The only difference between this and the input I provided you is a newline character after the function signature.

Sign up or log in to comment