You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: examples/reranker/README.md
+7-2Lines changed: 7 additions & 2 deletions
Original file line number
Diff line number
Diff line change
@@ -13,7 +13,8 @@ This implementation adds a new pooling type `LLAMA_POOLING_TYPE_RANK` which enab
13
13
## Installation
14
14
15
15
```bash
16
-
# Clone the repository
16
+
# Follow instructions to clone repo.
17
+
# Navigate to examples reranker
17
18
cd examples/reranker
18
19
19
20
# Build the project
@@ -67,4 +68,8 @@ Not tested others, but anything supported by llama.cpp should work.
67
68
68
69
## Implementation Details
69
70
70
-
This is a close Rust implementation of the reranker implementation discussed in [llama.cpp PR #9510](https://github.com/ggerganov/llama.cpp/pull/9510). Key features include:
71
+
This is a close Rust implementation of the reranker implementation discussed in [llama.cpp PR #9510](https://github.com/ggerganov/llama.cpp/pull/9510).
72
+
73
+
## Potential issues
74
+
75
+
The bos, eos, sep tokens are being hardcoded. We need to ideally get it from the model and build out the prompts based on each specific model.
0 commit comments