Edit model card

Built with Axolotl

ping98k/th-7b-20gb-base

This model is a continue pre-training version of openthaigpt/openthaigpt-1.0.0-beta-7b-chat-ckpt-hf on the 20GB Thai dataset. It achieves the following results on the evaluation set:

  • Loss: 1.5721

Inference with Pipeline

import torch
from transformers import pipeline
text_generator = pipeline("text-generation", model="ping98k/th-7b-20gb-base", torch_dtype=torch.bfloat16, device_map="auto")
print(text_generator("แบบจำลองทางวิทยาศาสตร์ (scientific modeling) คือ", max_length=50))

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.00015
  • train_batch_size: 4
  • eval_batch_size: 4
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 2
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 32
  • total_eval_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 100
  • num_epochs: 1

Training results

Training Loss Epoch Step Validation Loss
4.0347 0.0 1 4.0530
2.2753 0.05 1179 2.2083
2.1613 0.1 2358 2.0422
2.0696 0.15 3537 1.9526
1.945 0.2 4716 1.8886
1.6807 0.25 5895 1.8340
1.5838 0.3 7074 1.7961
1.7497 0.35 8253 1.7548
1.535 0.4 9432 1.7237
1.9632 0.45 10611 1.6878
1.9091 0.5 11790 1.6631
1.6837 0.55 12969 1.6344
1.7054 0.6 14148 1.6131
1.463 0.65 15327 1.5980
1.5538 0.7 16506 1.5853
1.5095 0.75 17685 1.5780
1.7322 0.8 18864 1.5742
1.5645 0.85 20043 1.5727
1.72 0.9 21222 1.5722
1.5882 0.95 22401 1.5721

Framework versions

  • Transformers 4.35.1
  • Pytorch 2.0.1+cu118
  • Datasets 2.14.7
  • Tokenizers 0.14.1
Downloads last month
12
Safetensors
Model size
6.94B params
Tensor type
BF16
·
Inference Examples
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.

Model tree for ping98k/th-7b-20gb-base

Finetuned
(7)
this model

Dataset used to train ping98k/th-7b-20gb-base