r5

.gitattributes

@@ -99,3 +99,5 @@ results/glm-4-9b_lora_sft_bf16-p1.csv filter=lfs diff=lfs merge=lfs -text
 results/llama3-8b_lora_sft_bf16-p1_en.csv filter=lfs diff=lfs merge=lfs -text
 results/mgtv-results_internlm_best.csv filter=lfs diff=lfs merge=lfs -text
 results/test_b-results_r4.csv filter=lfs diff=lfs merge=lfs -text
+results/mgtv-results_colab_p2_gemma2.csv filter=lfs diff=lfs merge=lfs -text
+results/test_b-results_r5.csv filter=lfs diff=lfs merge=lfs -text

results/test_b-results_r5.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af0267f312e20bb6538edf14702512a635b676dd099cd7e7209f6ee4a60ede2f
+size 23133371

scripts/test-mgtv.sh

@@ -24,17 +24,20 @@ export USING_LLAMA_FACTORY=false
 #export ADAPTER_NAME_OR_PATH=llama-factory/saves/internlm2_5_7b/lora/sft_bf16_p2_full_r3/checkpoint-140
 
 export MODEL_NAME=google/gemma-2-9b-it
-export ADAPTER_NAME_OR_PATH=outputs/checkpoint-6000
 export USING_P1_PROMPT_TEMPLATE=false
-
-#export MODEL_NAME=shenzhi-wang/Llama3-8B-Chinese-Chat
-#export ADAPTER_NAME_OR_PATH=llama-factory/saves/llama3-8b/lora/sft_bf16_p1_full_r4/checkpoint-140
-#export USING_P1_PROMPT_TEMPLATE=true
-
 export LOGICAL_REASONING_DATA_PATH=datasets/mgtv
 
-export TEST_DATA=test_b
-export LOGICAL_REASONING_RESULTS_PATH=results/$TEST_DATA-results_r5.csv
+export LOGICAL_REASONING_RESULTS_PATH=results/mgtv-results_colab_p2_gemma2.csv
+export ADAPTER_NAME_OR_PATH=outputs/checkpoint-7000
+echo "Eval $MODEL_NAME with $ADAPTER_NAME_OR_PATH"
+python llm_toolkit/eval_logical_reasoning.py
 
+
+export ADAPTER_NAME_OR_PATH=outputs/checkpoint-12000
 echo "Eval $MODEL_NAME with $ADAPTER_NAME_OR_PATH"
 python llm_toolkit/eval_logical_reasoning.py
+
+export TEST_DATA=test_b
+export LOGICAL_REASONING_RESULTS_PATH=results/$TEST_DATA-results_r6.csv
+echo "Test $MODEL_NAME with $ADAPTER_NAME_OR_PATH"
+python llm_toolkit/eval_logical_reasoning.py