metadata

license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - samsum
metrics:
  - rouge
model-index:
  - name: flan-t5-base
    results:
      - task:
          name: Summarization
          type: summarization
        dataset:
          name: samsum
          type: samsum
          split: validation
        metrics:
          - name: Rouge1
            type: rouge
            value: 46.819
pipeline_tag: summarization
inference: false
library_name: transformers
language:
  - en

flan-t5-base-cnn-samsum-lora

This model is a fine-tuned version of braindao/flan-t5-cnn on the samsum dataset.

The base model braindao/flan-t5-cnn is a fine-tuned verstion of google/flan-t5-base on the cnn_dailymail 3.0.0 dataset.

Check out sooolee/bart-large-cnn-finetuned-samsum-lora the model fine-tuned for the same purpose.

Model description

This model further finetuned braindao/flan-t5-cnn on the more conversational samsum dataset.
Huggingface PEFT Library LoRA (r = 16) and bitsandbytes int-8 was used to speed up training and reduce the model size.
Only 1.7M parameters were trained (0.71% of original flan-t5-base 250M parameters).
The model checkpoint is just 7MB.

Intended uses & limitations

Summarize transcripts such as YouTube transcripts.

Training and evaluation data

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.001
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 5

Training results

train_loss: 1.47
rogue1: 46.819522%
rouge2: 20.898074%
rougeL: 37.300937%
rougeLsum: 37.271341%

How to use

Note 'max_new_tokens=60' is used in the below example to control the length of the summary. FLAN-T5 model has max generation length = 200 and min generation length = 20 (default).

import torch
from peft import PeftModel, PeftConfig
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

# Load peft config for pre-trained checkpoint etc.
peft_model_id = "sooolee/flan-t5-base-cnn-samsum-lora"
config = PeftConfig.from_pretrained(peft_model_id)

# load base LLM model and tokenizer
model = AutoModelForSeq2SeqLM.from_pretrained(config.base_model_name_or_path, device_map='auto') # load_in_8bit=True, 
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)

# Load the Lora model
model = PeftModel.from_pretrained(model, peft_model_id, device_map='auto')

# Tokenize the text inputs
texts = "<e.g. Part of YouTube Transcript>"
inputs = tokenizer(texts, return_tensors="pt", padding=True, ) # truncation=True

# Make inferences
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
with torch.no_grad():    
    output = self.model.generate(input_ids=inputs["input_ids"].to(device), max_new_tokens=60, do_sample=True, top_p=0.9)
    summary = self.tokenizer.batch_decode(output.detach().cpu().numpy(), skip_special_tokens=True)

summary

Framework versions

Transformers 4.27.2
Pytorch 1.13.1+cu116
Datasets 2.9.0
Tokenizers 0.13.3