File size: 1,410 Bytes
23b59bf |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 |
---
license: mit
---
this is the placeholder model card for the (finess-benchmark-space)[https://huggingface.co/spaces/enzoescipy/finesse-benchmark-space] and its (database)[https://huggingface.co/datasets/enzoescipy/finesse-benchmark-results].
```python
import torch
from typing import List
from transformers import AutoConfig, PreTrainedModel # Optional: for loading configs
from finesse_benchmark.interfaces import FinesseSynthesizer
# --- Custom Embedder Example ---
# Uncomment and customize this class for your embedder.
class AverageSynthesizer(FinesseSynthesizer):
"""
Average Synthesizer: Computes the mean of input embeddings without using any model.
"""
def __init__(self, config_path: str):
super().__init__()
# No model to load for average pooling
print(f"{self.__class__.__name__} initialized - Average pooling ready.")
def synthesize(self, embeddings: torch.Tensor, **kwargs) -> torch.Tensor:
"""
Average synthesis: Compute the mean along the sequence dimension.
Args:
embeddings: torch.Tensor of shape (batch, seq_len, embedding_dim)
**kwargs: Additional arguments
Returns:
torch.Tensor of shape (batch, embedding_dim)
"""
return embeddings.mean(dim=1)
def device(self):
return "cpu"
```
it just averages the embedding vectors. |