Add application file

app.py

@@ -0,0 +1,62 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
+import gradio as gr
+from peft import PeftModel
+
+# Define the base model ID
+base_model_id = "meta-llama/Llama-2-13b-hf"
+
+# Create a BitsAndBytesConfig object with the corrected settings
+quantization_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_use_double_quant=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.bfloat16,
+    load_in_8bit_fp32_cpu_offload=True  # Set as suggested in the error
+)
+
+# Load the base model with the updated quantization configuration
+# Adjust 'device_map' based on your system's GPU configuration
+base_model = AutoModelForCausalLM.from_pretrained(
+    base_model_id,  
+    quantization_config=quantization_config,
+    trust_remote_code=True,
+    use_auth_token=True
+)
+
+# Load the tokenizer
+tokenizer = AutoTokenizer.from_pretrained(base_model_id, add_bos_token=True, trust_remote_code=True)
+
+# Load the fine-tuned model
+ft_model = PeftModel.from_pretrained(base_model, "checkpoint-2800")
+
+def formatting_func(job_description):
+    text = f"### The job description: {job_description}\n ### The skills: "
+    return text
+
+def generate_skills(job_description):
+    formatted_text = formatting_func(job_description)
+    model_input = tokenizer(formatted_text, return_tensors="pt").to("cuda")
+
+    ft_model.eval()
+    with torch.no_grad():
+        output_tokens = ft_model.generate(**model_input, max_new_tokens=200)[0]
+        generated_text = tokenizer.decode(output_tokens, skip_special_tokens=True)
+    
+    # Extract the text after "### The skills:" and before "### The qualifications:"
+    skills_start_index = generated_text.find("### The skills:") + len("### The skills:")
+    qualifications_start_index = generated_text.find("### The qualifications:")
+    
+    if qualifications_start_index != -1:
+        skills_text = generated_text[skills_start_index:qualifications_start_index].strip()
+    else:
+        skills_text = generated_text[skills_start_index:].strip()
+
+    return skills_text
+
+# Define the Gradio interface
+inputs = gr.Textbox(lines=10, label="Job description:", placeholder="Enter or paste the job description here...")
+outputs = gr.Textbox(label="Required skills:", placeholder="The required skills will be displayed here...")
+
+gr.Interface(fn=generate_skills, inputs=inputs, outputs=outputs, title="Job Skills Analysis",
+             description="Paste the job description in the text box below and the model will show the required skills for candidates.").launch()