alonzogarbanzo/Bloom-1b7-winograd-wsc-IT-baseline

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

Bloom-1b7-winograd-wsc-IT-baseline

This model is a fine-tuned version of bigscience/bloom-1b7 on an unknown dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

Instruction Tuned on the winograd-wsc task here: https://huggingface.co/datasets/adambjorn/UnrelatedForgettingOverhead/viewer/winograd_wsc

Training procedure

Given some prompts:

prompts = [
    "Determine which option the pronoun refers to in this text: ",
    "Given the text, identify the referent of the pronoun among these options: ",
    "Read the text and decide which option is referred to by the pronoun: ",
    "In the text below, to whom or what does the pronoun refer? Choose from the options: ",
]

Each example is concatenated with the prompt, text, pronoun, quote, options and correct option like so:

# Concatenate the selected prompt, text, pronoun, quote, options, and the correct option into a single string
input_text = f"{prompt}Text: '{text}' Pronoun: '{pronoun}', Quote: '{quote}'. {options_text}. {correct_option}. </s> "

Training hyperparameters

The following hyperparameters were used during training:

• —learning_rate: 5e-05
• —trainbatchsize: 1
• —evalbatchsize: 8
• —seed: 42
• —gradientaccumulationsteps: 4
• —totaltrainbatch_size: 4
• —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• —lrschedulertype: linear
• —num_epochs: 10
• —mixedprecisiontraining: Native AMP

Training results

Final results: {'loss': 0.0983, 'gradnorm': 4.820842266082764, 'learningrate': 6.000000000000001e-07, 'epoch': 10.0}

Average results: {'trainruntime': 452.2725, 'trainsamplespersecond': 4.422, 'trainstepspersecond': 1.106, 'trainloss': 0.33704672479629516, 'epoch': 10.0}

Framework versions

• —Transformers 4.38.1
• —Pytorch 2.2.0+cu121
• —Datasets 2.17.0
• —Tokenizers 0.15.2