Bases: EmbeddingModelOperationalMetric
Cache hit rate for embedding inputs: cached_tokens / prompt_tokens.
Example:
program.compile(
metrics=[
synalinks.metrics.EmbeddingCacheHitRate(),
],
)
Source code in synalinks/src/metrics/em_metrics.py
| @synalinks_export("synalinks.metrics.EmbeddingCacheHitRate")
class EmbeddingCacheHitRate(EmbeddingModelOperationalMetric):
"""Cache hit rate for embedding inputs: cached_tokens / prompt_tokens.
Example:
```python
program.compile(
metrics=[
synalinks.metrics.EmbeddingCacheHitRate(),
],
)
```
"""
def __init__(self, name="embedding_cache_hit_rate"):
super().__init__(name=name)
def result(self):
prompt = self._delta("prompt_tokens")
if prompt <= 0:
return 0.0
return self._delta("cached_tokens") / prompt
|