DEVCamiloSepulveda/2-LLAMA3SP-appceleratorstudio

1---
2license: llama3.2
3language:
4- en
5base_model: meta-llama/Llama-3.2-1B
6pipeline_tag: text-classification
7library_name: peft
8tags:
9- regression
10- story-point-estimation
11- software-engineering
12datasets:
13- appceleratorstudio
14metrics:
15- mae
16- mdae
17model-index:
18- name: llama-3.2-1b-story-point-estimation
19  results:
20  - task:
21      type: regression
22      name: Story Point Estimation
23    dataset:
24      name: appceleratorstudio Dataset
25      type: appceleratorstudio
26      split: test
27    metrics:
28    - type: mae
29      value: 1.551
30      name: Mean Absolute Error (MAE)
31    - type: mdae
32      value: 0.775
33      name: Median Absolute Error (MdAE)
34---
35# LLAMA 3 Story Point Estimator - appceleratorstudio36 37This model is fine-tuned on issue descriptions from appceleratorstudio and tested on appceleratorstudio for story point estimation.38 39## Model Details40- Base Model: LLAMA 3.2 1B41- Training Project: appceleratorstudio42- Test Project: appceleratorstudio43- Task: Story Point Estimation (Regression)44- Architecture: PEFT (LoRA)45- Tokenizer: SP Word Level46 47- Input: Issue titles48- Output: Story point estimation (continuous value)49 50## Usage51```python52from transformers import AutoModelForSequenceClassification53from peft import PeftConfig, PeftModel54from tokenizers import Tokenizer55 56# Load peft config model57config = PeftConfig.from_pretrained("DEVCamiloSepulveda/2-LLAMA3SP-appceleratorstudio")58 59# Load tokenizer and model60tokenizer = Tokenizer.from_pretrained("DEVCamiloSepulveda/2-LLAMA3SP-appceleratorstudio")61base_model = AutoModelForSequenceClassification.from_pretrained(62    config.base_model_name_or_path,63    num_labels=1,64    torch_dtype=torch.float16,65    device_map='auto'66)67model = PeftModel.from_pretrained(base_model, "DEVCamiloSepulveda/2-LLAMA3SP-appceleratorstudio")68 69# Prepare input text70text = "Your issue description here"71inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=20, padding="max_length")72 73# Get prediction74outputs = model(**inputs)75story_points = outputs.logits.item()76```77 78## Training Details79- Fine-tuning method: LoRA (Low-Rank Adaptation)80- Sequence length: 20 tokens81- Best training epoch: 0 / 20 epochs82- Batch size: 3283- Training time: 68.385 seconds84- Mean Absolute Error (MAE): 1.55185- Median Absolute Error (MdAE): 0.77586### Framework versions87 88- PEFT 0.14.0