evaluation/prometheus_evaluation_analysis.py

+import pandas as pd
+import matplotlib.pyplot as plt
+import glob
+
+# Adjust the file pattern/path as needed
+csv_files = sorted(glob.glob("eval_reports/llama-finetuned-responses-evaluation-pass*.csv"))[:10]
+print(csv_files)
+# Create a subplot grid: 2 rows x 5 columns for 10 plots
+fig, axes = plt.subplots(nrows=5, ncols=2, figsize=(30, 30))
+axes = axes.flatten()
+
+for i, csv_file in enumerate(csv_files):
+    # Read the CSV with the first row as header.
+    df = pd.read_csv(csv_file)
+    
+    # Ensure the 'score' column is numeric; invalid parsing will become NaN.
+    df['score'] = pd.to_numeric(df['score'], errors='coerce')
+    
+    scores = df['score']
+    
+    # Calculate mean and mode
+    mean_score = scores.mean()
+    mode_series = scores.mode()
+    mode_scores = mode_series.tolist() if not mode_series.empty else []
+    # Create a string of modes separated by commas for display
+    mode_str = ", ".join(str(m) for m in mode_scores) if mode_scores else "None"
+    
+    # Create a dictionary for score counts
+    score_counts = scores.value_counts().sort_index().to_dict()
+
+    print(f'For {csv_file} the statistics are as follows:\nMean: {mean_score} ; Mode: {mode_str}; Score Frequency: {score_counts}' )
+    
+    # Plot: Bar plot for score vs frequency
+    ax = axes[i]
+    ax.bar(list(score_counts.keys()), list(score_counts.values()),
+           color='skyblue', edgecolor='black')
+    ax.set_title(f"Mean: {mean_score:.2f}, Mode: {mode_str}", fontsize=40)
+    ax.set_xlabel("Score", fontsize=34)
+    ax.set_ylabel("Frequency", fontsize=34)
+    ax.tick_params(axis='x', rotation=45, labelsize=34) 
+    ax.tick_params(axis='y', labelsize=34) 
+
+plt.tight_layout()
+plt.savefig("Prometheus Evaluation Results - Llama Finetuned Model - spaced.pdf")
+plt.show()

evaluation/slurm_cosine_simiarity.sh

+#!/usr/bin/zsh
+
+### Add basic configuration for job
+
+#SBATCH --job-name=cosine_simiarity_evaluation
+#SBATCH --output=logs/cosine_simiarity_evaluation_analysis%j.log
+#SBATCH --error=logs/cosine_simiarity_evaluation_analysis_error_%j.log
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=1
+#SBATCH --time=00:30:00
+
+
+###------------------------------------------------------------------------------------------------------------------------------
+
+### Run the project in work directory of the cluster (configure based on need!! 
+### RWTH File System : https://help.itc.rwth-aachen.de/en/service/rhr4fjjutttf/article/da307ec2c60940b29bd42ac483fc3ea7/
+cd $HPCWORK
+cd codebud/evaluation
+###------------------------------------------------------------------------------------------------------------------------------
+
+### JOB SCRIPT RUN
+module load GCCcore/.13.2.0
+module load Python/3.11.5
+module load CUDA
+
+source ../../venvs/codebud/bin/activate
+echo $VIRTUAL_ENV
+
+python --version
+
+python cosine_similarity_evaluation.py
+module unload CUDA
+module unload Python/3.11.5
+
+deactivate
+echo "Script ran successfully"
\ No newline at end of file

evaluation/slurm_evaluation_absoulte.sh

+#!/usr/bin/zsh
+
+### Add basic configuration for job
+#SBATCH --account=rwth1776
+#SBATCH --job-name=prometheus_evaluation
+#SBATCH --output=logs/prometheus_evaluation_%j.log
+#SBATCH --error=logs/prometheus_evaluation_error_%j.log
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=5
+#SBATCH --gres=gpu:1
+#SBATCH --time=3:00:00
+
+
+###------------------------------------------------------------------------------------------------------------------------------
+
+### Run the project in work directory of the cluster (configure based on need!! 
+### RWTH File System : https://help.itc.rwth-aachen.de/en/service/rhr4fjjutttf/article/da307ec2c60940b29bd42ac483fc3ea7/
+cd $HPCWORK
+cd codebud/evaluation
+###------------------------------------------------------------------------------------------------------------------------------
+
+### JOB SCRIPT RUN
+module load GCCcore/.13.2.0
+module load Python/3.11.5
+module load CUDA
+
+source ../../venvs/codebud/bin/activate
+echo $VIRTUAL_ENV
+
+python --version
+
+python codebud_prometheus_absoulte_eval.py
+
+module unload CUDA
+module unload Python/3.11.5
+
+deactivate
+echo "Script ran successfully"
\ No newline at end of file

evaluation/slurm_prometheus_evaluation_analysis.sh

+#!/usr/bin/zsh
+
+### Add basic configuration for job
+
+#SBATCH --job-name=prometheus_evaluation
+#SBATCH --output=logs/prometheus_evaluation_analysis%j.log
+#SBATCH --error=logs/prometheus_evaluation_analysis_error_%j.log
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=1
+#SBATCH --time=00:10:00
+
+
+###------------------------------------------------------------------------------------------------------------------------------
+
+### Run the project in work directory of the cluster (configure based on need!! 
+### RWTH File System : https://help.itc.rwth-aachen.de/en/service/rhr4fjjutttf/article/da307ec2c60940b29bd42ac483fc3ea7/
+cd $HPCWORK
+cd codebud/evaluation
+###------------------------------------------------------------------------------------------------------------------------------
+
+### JOB SCRIPT RUN
+module load GCCcore/.13.2.0
+module load Python/3.11.5
+module load CUDA
+
+source ../../venvs/codebud/bin/activate
+echo $VIRTUAL_ENV
+
+python --version
+
+python prometheus_evaluation_analysis.py
+module unload CUDA
+module unload Python/3.11.5
+
+deactivate
+echo "Script ran successfully"
\ No newline at end of file

fine-tuning/main_llama_fine_tuning.py

+from unsloth import FastLanguageModel, is_bfloat16_supported
+from unsloth.chat_templates import get_chat_template, train_on_responses_only
+from datasets import load_dataset
+from trl import SFTTrainer
+from transformers import TrainingArguments, DataCollatorForSeq2Seq
+import wandb
+import torch
+import datetime
+
+print(f'Started the script at {datetime.datetime.now()}', flush=True)
+max_seq_length = 32768 # Choose any! We auto support RoPE Scaling internally!
+dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
+load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+    model_name = "unsloth/Meta-Llama-3.1-8B-Instruct",
+    max_seq_length = max_seq_length,
+    dtype = dtype,
+    load_in_4bit = load_in_4bit,
+)
+print(f'Model loaded successfully at {datetime.datetime.now()}', flush=True)
+
+model = FastLanguageModel.get_peft_model(
+    model,
+    r = 64, # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
+    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
+                      "gate_proj", "up_proj", "down_proj",],
+    lora_alpha = 64,
+    lora_dropout = 0, # Supports any, but = 0 is optimized
+    bias = "none",    # Supports any, but = "none" is optimized
+    # [NEW] "unsloth" uses 30% less VRAM, fits 2x larger batch sizes!
+    use_gradient_checkpointing = "unsloth", # True or "unsloth" for very long context
+    random_state = 3407,
+    use_rslora = False,  # We support rank stabilized LoRA
+    loftq_config = None, # And LoftQ
+)
+
+tokenizer = get_chat_template(
+    tokenizer,
+    chat_template = "llama-3.1",
+)
+
+dataset = load_dataset("atharva2721/llama-standardized-refined-train-aggregated", split = "train")
+validation_dataset = load_dataset("atharva2721/llama-standardized-refined-test-aggregated", split = "train")
+
+wandb.init(project="codebud")
+
+trainer = SFTTrainer(
+    model = model,
+    tokenizer = tokenizer,
+    train_dataset = dataset,
+    eval_dataset=validation_dataset,
+    dataset_text_field = "text",
+    max_seq_length = max_seq_length,
+    data_collator = DataCollatorForSeq2Seq(tokenizer = tokenizer),
+    dataset_num_proc = 4,
+    packing = False, # Can make training 5x faster for short sequences.
+    args = TrainingArguments(
+        per_device_train_batch_size = 1,
+        gradient_accumulation_steps = 4, # Fixed major bug in latest Unsloth
+        warmup_ratio = 0.1,
+        num_train_epochs = 3, # Set this for 1 full training run.
+        #max_steps = 60,
+        learning_rate = 2e-5,
+        fp16 = not is_bfloat16_supported(),
+        bf16 = is_bfloat16_supported(),
+        eval_strategy="steps",
+        eval_steps=410,
+        per_device_eval_batch_size = 1,
+        fp16_full_eval = not is_bfloat16_supported(),
+        bf16_full_eval = is_bfloat16_supported(),
+        logging_steps = 10,
+        save_steps = 410,
+        optim = "paged_adamw_8bit", # Save more memory
+        weight_decay = 0.01,
+        lr_scheduler_type = "cosine",
+        seed = 3407,
+        output_dir = "outputs",
+        report_to = "wandb", # Use this for WandB etc
+        run_name = "run-name"
+    ),
+)
+
+trainer = train_on_responses_only(
+    trainer,
+    instruction_part = "<|start_header_id|>user<|end_header_id|>\n\n",
+    response_part = "<|start_header_id|>assistant<|end_header_id|>\n\n",
+)
+
+tokenizer.decode(trainer.train_dataset[0]["input_ids"])
+
+space = tokenizer(" ", add_special_tokens = False).input_ids[0]
+tokenizer.decode([space if x == -100 else x for x in trainer.train_dataset[0]["labels"]])
+
+gpu_stats = torch.cuda.get_device_properties(0)
+start_gpu_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
+max_memory = round(gpu_stats.total_memory / 1024 / 1024 / 1024, 3)
+print(f"GPU = {gpu_stats.name}. Max memory = {max_memory} GB.", flush=True)
+print(f"{start_gpu_memory} GB of memory reserved.", flush=True)
+print(f'Everything initialized. Starting the training at {datetime.datetime.now()}', flush=True)
+
+trainer_stats = trainer.train()
+print(f'Successfully completed training at {datetime.datetime.now()}', flush=True)
+
+used_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
+used_memory_for_lora = round(used_memory - start_gpu_memory, 3)
+used_percentage = round(used_memory / max_memory * 100, 3)
+lora_percentage = round(used_memory_for_lora / max_memory * 100, 3)
+print(f"{trainer_stats.metrics['train_runtime']} seconds used for training.")
+print(
+    f"{round(trainer_stats.metrics['train_runtime']/60, 2)} minutes used for training."
+)
+print(f"Peak reserved memory = {used_memory} GB.")
+print(f"Peak reserved memory for training = {used_memory_for_lora} GB.")
+print(f"Peak reserved memory % of max memory = {used_percentage} %.")
+print(f"Peak reserved memory for training % of max memory = {lora_percentage} %.")
+
+print(f'Pushing model and tokenizer at {datetime.datetime.now()}', flush=True)
+model.save_pretrained("models/llama_finetuned_model")  # Local saving
+tokenizer.save_pretrained("models/llama_finetuned_model")
+model.push_to_hub("llama_finetuned_model") # Online saving
+tokenizer.push_to_hub("llama_finetuned_model") # Online saving
+
+wandb.finish()
+print(f'Run complete at {datetime.datetime.now()}', flush=True)
\ No newline at end of file

fine-tuning/main_qwen_fine_tuning.py

+from unsloth import FastLanguageModel, is_bfloat16_supported
+from unsloth.chat_templates import get_chat_template, train_on_responses_only
+from datasets import load_dataset
+from trl import SFTTrainer
+from transformers import TrainingArguments, DataCollatorForSeq2Seq
+import wandb
+import torch
+import datetime
+
+print(f'Started the script at {datetime.datetime.now()}', flush=True)
+max_seq_length = 32768 # Choose any! We auto support RoPE Scaling internally!
+dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
+load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+    model_name = "unsloth/Qwen2.5-Coder-7B-Instruct",
+    max_seq_length = max_seq_length,
+    dtype = dtype,
+    load_in_4bit = load_in_4bit,
+)
+print(f'Model loaded successfully at {datetime.datetime.now()}', flush=True)
+
+model = FastLanguageModel.get_peft_model(
+    model,
+    r = 64, # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128
+    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
+                      "gate_proj", "up_proj", "down_proj",],
+    lora_alpha = 64,
+    lora_dropout = 0, # Supports any, but = 0 is optimized
+    bias = "none",    # Supports any, but = "none" is optimized
+    # [NEW] "unsloth" uses 30% less VRAM, fits 2x larger batch sizes!
+    use_gradient_checkpointing = "unsloth", # True or "unsloth" for very long context
+    random_state = 3407,
+    use_rslora = False,  # We support rank stabilized LoRA
+    loftq_config = None, # And LoftQ
+)
+
+tokenizer = get_chat_template(
+    tokenizer,
+    chat_template = "qwen-2.5",
+)
+
+dataset = load_dataset("atharva2721/standardized-refined-train-aggregated", split = "train")
+validation_dataset = load_dataset("atharva2721/standardized-refined-val-test-aggregated", split = "train")
+
+wandb.init(project="codebud")
+
+trainer = SFTTrainer(
+    model = model,
+    tokenizer = tokenizer,
+    train_dataset = dataset,
+    eval_dataset=validation_dataset,
+    dataset_text_field = "text",
+    max_seq_length = max_seq_length,
+    data_collator = DataCollatorForSeq2Seq(tokenizer = tokenizer),
+    dataset_num_proc = 4,
+    packing = False, # Can make training 5x faster for short sequences.
+    args = TrainingArguments(
+        per_device_train_batch_size = 1,
+        gradient_accumulation_steps = 4, # Fixed major bug in latest Unsloth
+        warmup_ratio = 0.1,
+        num_train_epochs = 3, # Set this for 1 full training run.
+        #max_steps = 60,
+        learning_rate = 2e-5,
+        fp16 = not is_bfloat16_supported(),
+        bf16 = is_bfloat16_supported(),
+        eval_strategy="steps",
+        eval_steps=410,
+        per_device_eval_batch_size = 1,
+        fp16_full_eval = not is_bfloat16_supported(),
+        bf16_full_eval = is_bfloat16_supported(),
+        logging_steps = 10,
+        save_steps = 410,
+        optim = "paged_adamw_8bit", # Save more memory
+        weight_decay = 0.01,
+        lr_scheduler_type = "cosine",
+        seed = 3407,
+        output_dir = "outputs",
+        report_to = "wandb", # Use this for WandB etc
+        run_name = "run-name"
+    ),
+)
+
+trainer = train_on_responses_only(
+    trainer,
+    instruction_part = "<|im_start|>user\n",
+    response_part = "<|im_start|>assistant\n",
+)
+
+
+tokenizer.decode(trainer.train_dataset[0]["input_ids"])
+
+space = tokenizer(" ", add_special_tokens = False).input_ids[0]
+tokenizer.decode([space if x == -100 else x for x in trainer.train_dataset[0]["labels"]])
+
+gpu_stats = torch.cuda.get_device_properties(0)
+start_gpu_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
+max_memory = round(gpu_stats.total_memory / 1024 / 1024 / 1024, 3)
+print(f"GPU = {gpu_stats.name}. Max memory = {max_memory} GB.", flush=True)
+print(f"{start_gpu_memory} GB of memory reserved.", flush=True)
+print(f'Everything initialized. Starting the training at {datetime.datetime.now()}', flush=True)
+
+trainer_stats = trainer.train()
+print(f'Successfully completed training at {datetime.datetime.now()}', flush=True)
+
+used_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
+used_memory_for_lora = round(used_memory - start_gpu_memory, 3)
+used_percentage = round(used_memory / max_memory * 100, 3)
+lora_percentage = round(used_memory_for_lora / max_memory * 100, 3)
+print(f"{trainer_stats.metrics['train_runtime']} seconds used for training.")
+print(
+    f"{round(trainer_stats.metrics['train_runtime']/60, 2)} minutes used for training."
+)
+print(f"Peak reserved memory = {used_memory} GB.")
+print(f"Peak reserved memory for training = {used_memory_for_lora} GB.")
+print(f"Peak reserved memory % of max memory = {used_percentage} %.")
+print(f"Peak reserved memory for training % of max memory = {lora_percentage} %.")
+
+print(f'Pushing model and tokenizer at {datetime.datetime.now()}', flush=True)
+model.save_pretrained("models/finetuned_model_with_three_epochs_eval")  # Local saving
+tokenizer.save_pretrained("models/finetuned_model_with_three_epochs_eval")
+model.push_to_hub("finetuned_model_with_three_epochs_eval") # Online saving
+tokenizer.push_to_hub("finetuned_model_with_three_epochs_eval") # Online saving
+
+wandb.finish()
+print(f'Run complete at {datetime.datetime.now()}', flush=True)
\ No newline at end of file

fine-tuning/slurm_finetune.sh

+#!/usr/bin/zsh
+
+### Add basic configuration for job
+
+#SBATCH --job-name=fine_tuning
+#SBATCH --output=logs/fine_tuning_%j.log
+#SBATCH --error=logs/fine_tuning_error_%j.log
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=5
+#SBATCH --gres=gpu:1
+#SBATCH --time=22:00:00
+
+
+###------------------------------------------------------------------------------------------------------------------------------
+
+### Run the project in work directory of the cluster (configure based on need!! 
+### RWTH File System : https://help.itc.rwth-aachen.de/en/service/rhr4fjjutttf/article/da307ec2c60940b29bd42ac483fc3ea7/
+cd $HPCWORK
+cd codebud/fine-tuning
+###------------------------------------------------------------------------------------------------------------------------------
+
+### JOB SCRIPT RUN
+module load GCCcore/.13.2.0
+module load Python/3.11.5
+module load CUDA
+
+source ../../venvs/codebud/bin/activate
+echo $VIRTUAL_ENV
+
+python --version
+
+python main_llama_fine_tuning.py
+
+module unload CUDA
+module unload Python/3.11.5
+
+deactivate
+echo "Script ran successfully"
\ No newline at end of file

inference/llama_finetuned_inference.py

+from datasets import load_dataset, Dataset
+import re
+from unsloth import FastLanguageModel
+
+def data_generator(dataset):
+    for row in dataset:
+        yield row
+
+max_seq_length = 32768 # Choose any! We auto support RoPE Scaling internally!
+dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
+load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+    #model_name = "atharva2721/llama_finetuned_model",
+    model_name = "unsloth/Meta-Llama-3.1-8B-Instruct",
+    max_seq_length = max_seq_length,
+    dtype = dtype,
+    load_in_4bit = load_in_4bit,
+)
+FastLanguageModel.for_inference(model) # Enable native 2x faster inference
+
+reference_dataset = load_dataset("atharva2721/llama_inference_output", split="train", trust_remote_code=True)
+inference_output = []
+code_no = 0
+inferred_no = 0
+failed_no = 0
+
+for example in reference_dataset:
+    code_no += 1
+    content = f'''
+    Refine the C# code enclosed within tags [C#] and [/C#]. 
+    Provide the refined code must be enclosed within tags [refined_C#] and [/refined_C#]
+    The summary of changes must be enclosed within tags [code_changes] and [/code_changes].
+    
+    [C#]
+    {example["code"]}
+    [/C#]
+    '''
+    messages = [
+        {"role": "user", "content": content},
+    ]
+    inputs = tokenizer.apply_chat_template(
+        messages,
+        tokenize = True,
+        add_generation_prompt = True, # Must add for generation
+        return_tensors = "pt",
+    ).to("cuda")
+
+    should_retry = True
+    retry_no = 0
+    while should_retry:
+        print(f'Trying {code_no} for {retry_no} time', flush=True)
+        output_tensor = model.generate(input_ids = inputs,
+                        max_length  = max_seq_length,
+                        temperature = 0.6,
+                        repetition_penalty = 1.1
+                        )
+        
+        decoded = tokenizer.batch_decode(output_tensor)
+
+        output = ""
+        for text in decoded:
+            output += text
+        
+        output = output.split('<|start_header_id|>assistant<|end_header_id|>')
+        if len(output) == 2:
+            output = output[1]
+            
+            code_pattern = r'\[refined_C#\](.*?)\[/refined_C#\]'
+            summary_pattern = r'\[code_changes\](.*?)\[/code_changes\]'
+            
+            code_matches = re.search(code_pattern, output, re.DOTALL)
+            summary_matches = re.search(summary_pattern, output, re.DOTALL)
+
+            if code_matches and summary_matches:
+                refined_code = code_matches.group(1)
+                summary = summary_matches.group(1)
+                inference_output.append({'code': example["code"], 'base inference': refined_code, 'base summary' : summary, 'finetuned inference':example["finetuned inference"], 'finetuned summary': example["finetuned summary"], 'reference inference': example["reference inference"], 'reference summary': example["reference summary"]})
+                print(f'Code no. {code_no} refined successfully', flush=True)
+                should_retry = False
+                inferred_no += 1
+            
+        if retry_no == 2:
+            should_retry = False
+            print(f'Failed to refine code at {code_no}. Final try output: \n [failed_output]{output}[/failed_output]', flush=True)
+            failed_no +=1
+
+        retry_no += 1
+
+new_dataset = Dataset.from_generator(data_generator, gen_kwargs={"dataset": inference_output})
+new_dataset.push_to_hub('llama_inference_output_complete')
+print(f'Created and pushed total of {inferred_no} examples from total of {code_no} codes. Total failed inferences are {failed_no}', flush=True)
\ No newline at end of file

inference/llama_finetuned_inference_java.py

+from unsloth import FastLanguageModel
+max_seq_length = 32768 # Choose any! We auto support RoPE Scaling internally!
+dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
+load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+    #model_name = "atharva2721/llama_finetuned_model",
+    model_name = "unsloth/Meta-Llama-3.1-8B-Instruct",
+    max_seq_length = max_seq_length,
+    dtype = dtype,
+    load_in_4bit = load_in_4bit,
+)
+FastLanguageModel.for_inference(model) # Enable native 2x faster inference
+
+code1 = """
+package com.thealgorithms.recursion;
+
+// program to find power set of a string
+
+import java.util.ArrayList;
+import java.util.List;
+
+public final class genSubsets {
+
+    private genSubsets() {
+        throw new UnsupportedOperationException("Utility class");
+    }
+
+    public static List<String> subsetRecursion(String str) {
+        return doRecurr("", str);
+    }
+
+    private static List<String> doRecurr(String p, String up) {
+        if (up.isEmpty()) {
+            List<String> list = new ArrayList<>();
+            list.add(p);
+            return list;
+        }
+
+        // Taking the character
+        char ch = up.charAt(0);
+        // Adding the character in the recursion
+        List<String> left = doRecurr(p + ch, up.substring(1));
+        // Not adding the character in the recursion
+        List<String> right = doRecurr(p, up.substring(1));
+
+        left.addAll(right);
+
+        return left;
+    }
+}
+"""
+
+code2 = """
+using System;
+using System.Collections;
+namespace hub
+{
+    public class logicproxy
+    {
+        public logicproxy(juggle.Ichannel ch)
+        {
+            _hub_call_logic = new caller.hub_call_logic(ch);
+        }
+        public void reg_logic_sucess_and_notify_hub_nominate()
+        {
+            _hub_call_logic.reg_logic_sucess_and_notify_hub_nominate(hub.name);
+        }
+        public void call_logic(String module_name, String func_name, params object[] argvs)
+        {
+            ArrayList _argvs = new ArrayList();
+            foreach (var o in argvs)
+            {
+                _argvs.Add(o);
+            }
+            _hub_call_logic.hub_call_logic_mothed(module_name, func_name, _argvs);
+        }
+        private caller.hub_call_logic _hub_call_logic;
+    }
+}
+"""
+
+code3 = """
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+namespace Max_Number
+{
+    class Program
+    {
+        static void Main(string[] args)
+        {
+            int number = int.Parse(Console.ReadLine());
+            double max = double.Parse(Console.ReadLine());
+            for (int i = 1; i < number; i++)
+            {
+                double d = double.Parse(Console.ReadLine());
+                if (d > max)
+                    max = d;
+            }
+            Console.WriteLine(max);
+        }
+    }
+}
+"""
+
+code4 = """
+using System.Collections.Generic;
+using System.Linq;
+namespace Mirage.Urbanization.Simulation
+{
+    public class CityCategoryDefinition
+    {
+        public CityCategoryDefinition(string name, int minimumPopulation)
+        {
+            Name = name;
+            MinimumPopulation = minimumPopulation;
+        }
+        public string Name { get; }
+        public int MinimumPopulation { get; }
+        public static CityCategoryDefinition GetForPopulation(int population)
+        {
+            return Definitions
+                .Where(x => x.MinimumPopulation <= population)
+                .OrderByDescending(x => x.MinimumPopulation)
+                .First();
+        }
+        public static CityCategoryDefinition Village = new CityCategoryDefinition("Village", 0);
+        private static readonly IReadOnlyCollection<CityCategoryDefinition> Definitions = new[]
+        {
+            Village,
+            new CityCategoryDefinition("Town", 2000),
+            new CityCategoryDefinition("City", 10000),
+            new CityCategoryDefinition("Capital", 50000),
+            new CityCategoryDefinition("Metropolis", 100000)
+        }.ToList();
+    }
+}
+"""
+
+code5 = """
+using System;
+using System.Collections.Generic;
+using System.ComponentModel;
+using System.Data;
+using System.Drawing;
+using System.Drawing.Imaging;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+using System.Windows.Forms;
+namespace TagUIWordAddIn
+{
+    public partial class SnapshotBar : Form
+    {
+        public SnapshotBar()
+        {
+            InitializeComponent();
+        }
+        private void button1_Click(object sender, EventArgs e)
+        {
+            this.Hide();
+            Form1 f1 = new Form1();
+            if (checkBoxDelay.Checked)
+            {
+                System.Threading.Thread.Sleep(5000);
+                Form2 f2 = new Form2();
+                f1.Owner = f2;
+                f2.Show();
+                f1.Show();
+                f1.Closed += (s, args) =>
+                {
+                    this.Close();
+                    f2.Close();
+                };
+            }
+            else
+            {
+                f1.Show();
+                f1.Closed += (s, args) =>
+                {
+                    this.Close();
+                };
+            }
+            f1.Closed += (s, args) =>
+            {
+                this.Close();
+            };
+        }
+    }
+}
+"""
+
+
+codes = [code1]
+
+for code in codes:
+    print("******************Start of Generation******************")
+    content = f'''
+    Refine the java code enclosed within tags [java] and [/java]. 
+    Provide the refined code enclosed within tags [refined_java] and [/refined_java] and summary of changes enclosed within tags [code_changes] and [/code_changes].
+    
+    [java]
+    {code}
+    [/java]
+    '''
+
+    messages = [
+        {"role": "user", "content": content},
+    ]
+    inputs = tokenizer.apply_chat_template(
+        messages,
+        tokenize = True,
+        add_generation_prompt = True, # Must add for generation
+        return_tensors = "pt",
+    ).to("cuda")
+
+    from transformers import TextStreamer
+    text_streamer = TextStreamer(tokenizer, skip_prompt = True)
+    _ = model.generate(input_ids = inputs, streamer = text_streamer, max_new_tokens = 10000, temperature = 0.6)
+    
+    print("******************End of Generation******************")
\ No newline at end of file

inference/qwen_finetuned_inference.py

+from datasets import load_dataset, Dataset
+import re
+from unsloth import FastLanguageModel
+
+def data_generator(dataset):
+    for row in dataset:
+        yield row
+
+max_seq_length = 32768 
+dtype = None 
+load_in_4bit = True 
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+    #model_name = "atharva2721/qwen_finetuned_model",
+    model_name = "unsloth/Qwen2.5-Coder-7B-Instruct",
+    max_seq_length = max_seq_length,
+    dtype = dtype,
+    load_in_4bit = load_in_4bit,
+)
+FastLanguageModel.for_inference(model) 
+
+reference_dataset = load_dataset("atharva2721/qwen_inference_output", split="train", trust_remote_code=True)
+inference_output = []
+code_no = 0
+inferred_no = 0
+failed_no = 0
+
+for example in reference_dataset:
+    code_no += 1
+    content = f'''
+    Refine the C# code enclosed within tags [C#] and [/C#]. 
+    Provide the refined code enclosed within tags [refined_C#] and [/refined_C#]
+    The summary of changes must be enclosed within tags [code_changes] and [/code_changes].
+    
+    [C#]
+    {example["code"]}
+    [/C#]
+    '''
+    messages = [
+        {"role": "user", "content": content},
+    ]
+    inputs = tokenizer.apply_chat_template(
+        messages,
+        tokenize = True,
+        add_generation_prompt = True, # Must add for generation
+        return_tensors = "pt",
+    ).to("cuda")
+
+    should_retry = True
+    retry_no = 0
+    while should_retry:
+        output_tensor = model.generate(input_ids = inputs,
+                        max_length  = max_seq_length,
+                        temperature = 0.7
+                        )
+        
+        decoded = tokenizer.batch_decode(output_tensor)
+
+        output = ""
+        for text in decoded:
+            output += text
+        
+        output = output.split('<|im_start|>assistant')
+        if len(output) == 2:
+            output = output[1]
+            
+            code_pattern = r'\[refined_C#\](.*?)\[/refined_C#\]'
+            summary_pattern = r'\[code_changes\](.*?)\[/code_changes\]'
+            
+            code_matches = re.search(code_pattern, output, re.DOTALL)
+            summary_matches = re.search(summary_pattern, output, re.DOTALL)
+
+            if code_matches and summary_matches:
+                refined_code = code_matches.group(1)
+                summary = summary_matches.group(1)
+                inference_output.append({'code': example["code"], 'base inference':refined_code, 'base summary': summary,'finetuned inference': example["finetuned inference"], 'finetuned summary': example["finetuned summary"], 'reference inference': example["reference inference"], 'reference summary': example["reference summary"]})
+                print(f'Code no. {code_no} refined successfully', flush=True)
+                should_retry = False
+                inferred_no += 1
+            
+        if retry_no == 2:
+            should_retry = False
+            print(f'Failed to refine code at {code_no}. Final try output: \n [failed_output]{output}[/failed_output]', flush=True)
+            failed_no +=1
+
+        retry_no += 1
+
+new_dataset = Dataset.from_generator(data_generator, gen_kwargs={"dataset": inference_output})
+new_dataset.push_to_hub('qwen_inference_output_complete')
+print(f'Created and pushed total of {inferred_no} examples from total of {code_no} codes. Total failed inferences are {failed_no}', flush=True)
\ No newline at end of file

inference/qwen_finetuned_inference_java.py

+from unsloth import FastLanguageModel
+max_seq_length = 32768 # Choose any! We auto support RoPE Scaling internally!
+dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
+load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+    #model_name = "atharva2721/qwen_finetuned_model",
+    model_name = "unsloth/Qwen2.5-Coder-7B-Instruct",
+    max_seq_length = max_seq_length,
+    dtype = dtype,
+    load_in_4bit = load_in_4bit,
+)
+FastLanguageModel.for_inference(model) # Enable native 2x faster inference
+
+code1 = """
+package com.thealgorithms.recursion;
+
+// program to find power set of a string
+
+import java.util.ArrayList;
+import java.util.List;
+
+public final class genSubsets {
+
+    private genSubsets() {
+        throw new UnsupportedOperationException("Utility class");
+    }
+
+    public static List<String> subsetRecursion(String str) {
+        return doRecurr("", str);
+    }
+
+    private static List<String> doRecurr(String p, String up) {
+        if (up.isEmpty()) {
+            List<String> list = new ArrayList<>();
+            list.add(p);
+            return list;
+        }
+
+        // Taking the character
+        char ch = up.charAt(0);
+        // Adding the character in the recursion
+        List<String> left = doRecurr(p + ch, up.substring(1));
+        // Not adding the character in the recursion
+        List<String> right = doRecurr(p, up.substring(1));
+
+        left.addAll(right);
+
+        return left;
+    }
+}
+"""
+
+code2 = """
+using System;
+using System.Collections;
+namespace hub
+{
+    public class logicproxy
+    {
+        public logicproxy(juggle.Ichannel ch)
+        {
+            _hub_call_logic = new caller.hub_call_logic(ch);
+        }
+        public void reg_logic_sucess_and_notify_hub_nominate()
+        {
+            _hub_call_logic.reg_logic_sucess_and_notify_hub_nominate(hub.name);
+        }
+        public void call_logic(String module_name, String func_name, params object[] argvs)
+        {
+            ArrayList _argvs = new ArrayList();
+            foreach (var o in argvs)
+            {
+                _argvs.Add(o);
+            }
+            _hub_call_logic.hub_call_logic_mothed(module_name, func_name, _argvs);
+        }
+        private caller.hub_call_logic _hub_call_logic;
+    }
+}
+"""
+
+code3 = """
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+namespace Max_Number
+{
+    class Program
+    {
+        static void Main(string[] args)
+        {
+            int number = int.Parse(Console.ReadLine());
+            double max = double.Parse(Console.ReadLine());
+            for (int i = 1; i < number; i++)
+            {
+                double d = double.Parse(Console.ReadLine());
+                if (d > max)
+                    max = d;
+            }
+            Console.WriteLine(max);
+        }
+    }
+}
+"""
+
+code4 = """
+using System.Collections.Generic;
+using System.Linq;
+namespace Mirage.Urbanization.Simulation
+{
+    public class CityCategoryDefinition
+    {
+        public CityCategoryDefinition(string name, int minimumPopulation)
+        {
+            Name = name;
+            MinimumPopulation = minimumPopulation;
+        }
+        public string Name { get; }
+        public int MinimumPopulation { get; }
+        public static CityCategoryDefinition GetForPopulation(int population)
+        {
+            return Definitions
+                .Where(x => x.MinimumPopulation <= population)
+                .OrderByDescending(x => x.MinimumPopulation)
+                .First();
+        }
+        public static CityCategoryDefinition Village = new CityCategoryDefinition("Village", 0);
+        private static readonly IReadOnlyCollection<CityCategoryDefinition> Definitions = new[]
+        {
+            Village,
+            new CityCategoryDefinition("Town", 2000),
+            new CityCategoryDefinition("City", 10000),
+            new CityCategoryDefinition("Capital", 50000),
+            new CityCategoryDefinition("Metropolis", 100000)
+        }.ToList();
+    }
+}
+"""
+
+code5 = """
+using System;
+using System.Collections.Generic;
+using System.ComponentModel;
+using System.Data;
+using System.Drawing;
+using System.Drawing.Imaging;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+using System.Windows.Forms;
+namespace TagUIWordAddIn
+{
+    public partial class SnapshotBar : Form
+    {
+        public SnapshotBar()
+        {
+            InitializeComponent();
+        }
+        private void button1_Click(object sender, EventArgs e)
+        {
+            this.Hide();
+            Form1 f1 = new Form1();
+            if (checkBoxDelay.Checked)
+            {
+                System.Threading.Thread.Sleep(5000);
+                Form2 f2 = new Form2();
+                f1.Owner = f2;
+                f2.Show();
+                f1.Show();
+                f1.Closed += (s, args) =>
+                {
+                    this.Close();
+                    f2.Close();
+                };
+            }
+            else
+            {
+                f1.Show();
+                f1.Closed += (s, args) =>
+                {
+                    this.Close();
+                };
+            }
+            f1.Closed += (s, args) =>
+            {
+                this.Close();
+            };
+        }
+    }
+}
+"""
+
+
+codes = [code1]
+for code in codes:
+    print("******************Start of Generation******************")
+    content = f'''
+    Refine the java code enclosed within tags [java] and [/java]. 
+    Provide the refined code enclosed within tags [refined_java] and [/refined_java] and summary of changes enclosed within tags [code_changes] and [/code_changes].
+    
+    [java]
+    {code}
+    [/java]
+    '''
+    messages = [
+        {"role": "user", "content": content},
+    ]
+    inputs = tokenizer.apply_chat_template(
+        messages,
+        tokenize = True,
+        add_generation_prompt = True, # Must add for generation
+        return_tensors = "pt",
+    ).to("cuda")
+
+    from transformers import TextStreamer
+    text_streamer = TextStreamer(tokenizer, skip_prompt = True)
+    _ = model.generate(input_ids = inputs, 
+                    streamer = text_streamer, 
+                    max_new_tokens = 10000,
+                    temperature = 0.7
+                    )
+    
+    print("******************End of Generation******************")
\ No newline at end of file

inference/slurm_llama_inference.sh

+#!/usr/bin/zsh
+
+### Add basic configuration for job
+#SBATCH --account=rwth1776
+#SBATCH --job-name=llama_inference
+#SBATCH --output=logs/llama_inference_%j.log
+#SBATCH --error=logs/llama_inference_error%j.log
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=3
+#SBATCH --gres=gpu:1
+#SBATCH --time=20:00:00
+
+
+###------------------------------------------------------------------------------------------------------------------------------
+
+### Run the project in work directory of the cluster (configure based on need!! 
+### RWTH File System : https://help.itc.rwth-aachen.de/en/service/rhr4fjjutttf/article/da307ec2c60940b29bd42ac483fc3ea7/
+cd $HPCWORK
+cd codebud/inference
+###------------------------------------------------------------------------------------------------------------------------------
+
+### JOB SCRIPT RUN
+module load GCCcore/.13.2.0
+module load Python/3.11.5
+module load CUDA
+
+source ../../venvs/codebud/bin/activate
+echo $VIRTUAL_ENV
+
+python --version
+
+python llama_finetuned_inference.py
+
+module unload CUDA
+module unload Python/3.11.5
+
+deactivate
+echo "Script ran successfully"
\ No newline at end of file

inference/slurm_qwen_inference.sh

+#!/usr/bin/zsh
+
+### Add basic configuration for job
+
+#SBATCH --account=rwth1776
+#SBATCH --job-name=qwen_inference
+#SBATCH --output=logs/qwen_inference_%j.log
+#SBATCH --error=logs/qwen_inference_error_%j.log
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=3
+#SBATCH --gres=gpu:1
+#SBATCH --time=05:00:00
+
+
+###------------------------------------------------------------------------------------------------------------------------------
+
+### Run the project in work directory of the cluster (configure based on need!! 
+### RWTH File System : https://help.itc.rwth-aachen.de/en/service/rhr4fjjutttf/article/da307ec2c60940b29bd42ac483fc3ea7/
+cd $HPCWORK
+cd codebud/inference
+###------------------------------------------------------------------------------------------------------------------------------
+
+### JOB SCRIPT RUN
+module load GCCcore/.13.2.0
+module load Python/3.11.5
+module load CUDA
+
+source ../../venvs/codebud/bin/activate
+echo $VIRTUAL_ENV
+
+python --version
+
+#python qwen_finetuned_inference.py
+python qwen_finetuned_inference.py
+
+module unload CUDA
+module unload Python/3.11.5
+
+deactivate
+echo "Script ran successfully"
\ No newline at end of file

requirements.txt

+accelerate==1.1.1
+aiohappyeyeballs==2.4.3
+aiohttp==3.11.8
+aiosignal==1.3.1
+attrs==24.2.0
+certifi==2024.8.30
+charset-normalizer==3.4.0
+colorama==0.4.6
+datasets==3.1.0
+dill==0.3.8
+filelock==3.16.1
+frozenlist==1.5.0
+fsspec==2024.9.0
+huggingface-hub==0.26.3
+idna==3.10
+Jinja2==3.1.3
+MarkupSafe==2.1.5
+mpmath==1.3.0
+multidict==6.1.0
+multiprocess==0.70.16
+networkx==3.2.1
+numpy==2.1.3
+packaging==24.2
+pandas==2.2.3
+pillow==10.2.0
+propcache==0.2.0
+psutil==6.1.0
+pyarrow==18.1.0
+python-dateutil==2.9.0.post0
+pytz==2024.2
+PyYAML==6.0.2
+regex==2024.11.6
+requests==2.32.3
+safetensors==0.4.5
+setuptools==70.0.0
+six==1.16.0
+sympy==1.13.1
+tokenizers==0.20.3
+torch==2.5.1+cu118
+torchaudio==2.5.1+cu118
+torchvision==0.20.1+cu118
+tqdm==4.67.1
+transformers==4.46.3
+typing_extensions==4.12.2
+tzdata==2024.2
+urllib3==2.2.3
+xxhash==3.5.0
+yarl==1.18.0