626sirs/multimodal-sentiment-weights

Multimodal sentiment prediction checkpoints

This repository provides two frozen fine-tuned BERT branch checkpoints for a multimodal sentiment prediction system. The checkpoints are sanitized inference state dictionaries, not standalone Transformers AutoModel packages. Use the accompanying core inference code and configuration to load them.

The public base model is google-bert/bert-base-uncased (Apache-2.0). Exact assets and SHA256 checksums are listed below. No training data, labels, personal information, or optimizer state are included.

Checkpoints

• bert_seed260925.pt: SHA256 2e4fa0b965fc41494d4b65546adfede518de0282adf88628e25b394408f11e07

• bert_seed260926.pt: SHA256 6691e7adc1709ddec9b17c5563c0abe3d1833d58486868d939db049c8a9ad5f9

Pinned public base model

Revision: 86b5e0934494bd15c9632b12f734a8a67f723594

File: model.safetensors

SHA256: 68d45e234eb4a928074dfd868cead0219ab85354cc53d20e772753c6bb9169d3

Download pinned base weights