Update README with metrics
Browse files
README.md
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pretty_name: "/hub_data2/seohyun/saves/qwen3vl-2b-instruct_tuned/full/sft · happy8825/activitynet_validset results"
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- video-retrieval
|
| 7 |
+
- evaluation
|
| 8 |
+
- vllm
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# /hub_data2/seohyun/saves/qwen3vl-2b-instruct_tuned/full/sft · happy8825/activitynet_validset results
|
| 12 |
+
|
| 13 |
+
- **Model**: `/hub_data2/seohyun/saves/qwen3vl-2b-instruct_tuned/full/sft`
|
| 14 |
+
- **Dataset**: `happy8825/activitynet_validset`
|
| 15 |
+
- **Generated**: `2025-12-11 16:50:16Z`
|
| 16 |
+
|
| 17 |
+
## Metrics
|
| 18 |
+
| Metric | Value |
|
| 19 |
+
| --- | --- |
|
| 20 |
+
| Total samples | 4917 |
|
| 21 |
+
| With GT | 4917 |
|
| 22 |
+
| Parsed answers | 4913 |
|
| 23 |
+
| Top-1 accuracy | 0.449258 |
|
| 24 |
+
| Recall@5 | 0.449258 |
|
| 25 |
+
| MRR | 1 |
|
| 26 |
+
|
| 27 |
+
The uploaded JSON contains full per-sample predictions produced via `t3_infer_with_vllm.bash`.
|