Model Card: generate_reason
Model Name
generate_reason
Model Description
This model represents a fine-tuned version of the facebook/bart-large model, specifically adapted for the task of reason generator by annalysing resume with job description. The model has been trained to efficiently generate concise and relevant reason from extensive resume texts and JD. The fine-tuning process has tailored the original BART model to specialize in summarization tasks based on a specific dataset.
Model information
-Base Model: GebeyaTalent/generate_reason
-Finetuning Dataset: To be made available in the future.
Training Parameters
- Evaluation Strategy: epoch:
- Learning Rate: 5e-5
- Per Device Train Batch Size: 8:
- Per Device Eval Batch Size: 8
- Weight Decay: 0.01
- Save Total Limit: 5
- Number of Training Epochs: 5
- Predict with Generate: True
- Gradient Accumulation Steps: 1
- Optimizer: paged_adamw_32bit
- Learning Rate Scheduler Type: cosine
how to use
1. Install the transformers library:
pip install transformers
2. Import the necessary modules:
import torch
from transformers import BartTokenizer, BartForConditionalGeneration
3. Initialize the model and tokenizer:
model_name = 'GebeyaTalent/generate_reason'
tokenizer = BartTokenizer.from_pretrained(model_name)
model = BartForConditionalGeneration.from_pretrained(model_name)
4. Prepare the text to generate reason:
resume = 'your resume text here"
job_description = "your job_description here"
Concatenate the resume and job description with a delimiter
combined_text = "Resume: " + ["resume"] + " Job Description: " + ["job_description"]
inputs = tokenizer(combined_text, return_tensors="pt", truncation=True, padding="max_length", max_length=1024)
5. Generate reason
reason_ids = model.generate(inputs["input_ids"], num_beams=4, max_length=150, early_stopping=True)
reason = tokenizer.decode(reason_ids[0], skip_special_tokens=True)
6. Output the summary:
print("Reason:", reason)
Model Card Authors
Dereje Hinsermu
Model Card Contact
- Downloads last month
- 15
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.