upload eval example for mldr

examples/evaluation/mldr/eval_mldr.sh

@@ -0,0 +1,38 @@
+if [ -z "$HF_HUB_CACHE" ]; then
+    export HF_HUB_CACHE="$HOME/.cache/huggingface/hub"
+fi
+
+dataset_names="hi"
+
+eval_args="\
+    --eval_name mldr \
+    --dataset_dir /share/jianlv/evaluation/mldr/data \
+    --dataset_names $dataset_names \
+    --splits test \
+    --corpus_embd_save_dir /share/jianlv/evaluation/mldr/corpus_embd \
+    --output_dir /share/jianlv/evaluation/mldr/search_results \
+    --search_top_k 1000 --rerank_top_k 100 \
+    --cache_path $HF_HUB_CACHE \
+    --overwrite False \
+    --k_values 10 100 \
+    --eval_output_method markdown \
+    --eval_output_path /share/jianlv/evaluation/mldr/mldr_eval_results.md \
+    --eval_metrics ndcg_at_10 \
+"
+
+model_args="\
+    --embedder_name_or_path BAAI/bge-m3 \
+    --reranker_name_or_path BAAI/bge-reranker-v2-m3 \
+    --devices cuda:0 cuda:1 \
+    --cache_dir $HF_HUB_CACHE \
+    --embedder_passage_max_length 8192 \
+    --reranker_max_length 8192 \
+"
+
+cmd="python -m FlagEmbedding.evaluation.mldr \
+    $eval_args \
+    $model_args \
+"
+
+echo $cmd
+eval $cmd