BlenderBot Conversational Chatbot

Fine-tuned facebook/blenderbot-400M-distill on blended_skill_talk for open-domain multi-turn conversation.

Model Details

Property Value
Base model facebook/blenderbot-400M-distill
Parameters 364.8M
Dataset blended_skill_talk
Best Val PPL 14.16
Epochs 2 (early stopping)
Learning rate 2e-5

Usage

from transformers import BlenderbotTokenizer, BlenderbotForConditionalGeneration

tokenizer = BlenderbotTokenizer.from_pretrained("nilotpaldhar2004/blenderbot-chatbot")
model = BlenderbotForConditionalGeneration.from_pretrained("nilotpaldhar2004/blenderbot-chatbot")

inputs = tokenizer("Hello! How are you?", return_tensors="pt")
output = model.generate(**inputs, max_new_tokens=60, num_beams=2)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Training Details

  • Samples: 60,000 pairs from blended_skill_talk
  • Effective batch: 64 (batch 8 x grad accum 8)
  • Warmup: 15% of total steps
  • Early stopping patience: 2

Author

Nilotpal — CS student, AI/ML

Downloads last month
827
Safetensors
Model size
0.4B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for nilotpaldhar2004/blenderbot-chatbot

Finetuned
(23)
this model

Dataset used to train nilotpaldhar2004/blenderbot-chatbot

Space using nilotpaldhar2004/blenderbot-chatbot 1