YouTube comment sentiment (3-class)

DistilRoBERTa encoder (top layers fine-tuned) + custom head, CLS pooling. Labels: Negative (0), Neutral (1), Positive (2). Max length 128.

This is a custom nn.Module, not a stock AutoModelForSequenceClassification. Load it with model.py from this repo:

import json, torch
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file
from transformers import AutoTokenizer
from model import YTmodel

m = YTmodel(pretrained=False)
m.load_state_dict(load_file(hf_hub_download("Avinash159-159/yt-comment-sentiment", "model.safetensors")))
m.eval()

Live demo / API: https://huggingface.co/spaces/Avinash159-159/yt-comment-sentiment-demo