|
|
--- |
|
|
title: Ernie 4.5 Logprob Compare |
|
|
emoji: 🐢 |
|
|
colorFrom: blue |
|
|
colorTo: pink |
|
|
sdk: gradio |
|
|
sdk_version: 5.41.0 |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
license: mit |
|
|
short_description: Compare the output of ERNIE-4.5-0.3B Instruct and Base model |
|
|
--- |
|
|
|
|
|
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference |
|
|
|
|
|
# 🔍 ERNIE 4.5 LogProb Compare |
|
|
|
|
|
This demo compares two ERNIE 4.5 models — `baidu/ERNIE-4.5-0.3B-PT` and `baidu/ERNIE-4.5-0.3B-Base-PT` — on the same input sentence. |
|
|
|
|
|
For each token, the app displays: |
|
|
- Token-wise log probabilities |
|
|
- Top-k predicted tokens (with probability) |
|
|
- Total sentence log probability |
|
|
|
|
|
Useful for debugging, evaluation, or teaching language model internals. |
|
|
|
