nv4090 results

llm_toolkit/eval_logical_reasoning.py

@@ -94,7 +94,7 @@ if adapter_name_or_path is not None:
     model_name += "/" + adapter_name_or_path.split("/")[-1]
 
 save_results(
-    f"{model_name}_{dtype}{'_4bit' if load_in_4bit else ''}",
+    f"{model_name}_{dtype}{'_4bit' if load_in_4bit else ''}{'_lf' if using_llama_factory else ''}",
     results_path,
     datasets["test"],
     predictions,

llm_toolkit/llm_utils.py

@@ -88,7 +88,9 @@ def check_gpu():
     # If we have a GPU available, we'll set our device to GPU. We'll use this device variable later in our code.
     if is_cuda:
         device = torch.device("cuda")
-        print("GPU is available")
+        print("CUDA is available, we have found ", torch.cuda.device_count(), " GPU(s)")
+        print(torch.cuda.get_device_name(0))
+        print("CUDA version: " + torch.version.cuda)
     elif torch.backends.mps.is_available():
         device = torch.device("mps")
         print("MPS is available")

scripts/eval-mgtv-nv4090.sh

@@ -0,0 +1,35 @@
+#!/bin/sh
+
+BASEDIR=$(dirname "$0")
+cd $BASEDIR/..
+echo Current Directory:
+pwd
+
+BASEDIR=`pwd`
+
+nvidia-smi
+uname -a
+cat /etc/os-release
+lscpu
+grep MemTotal /proc/meminfo
+
+export LOAD_IN_4BIT=false
+
+export MODEL_NAME=internlm/internlm2_5-7b-chat-1m
+export ADAPTER_NAME_OR_PATH=inflaton-ai/InternLM_2_5-7b_LoRA-Adapter
+export LOGICAL_REASONING_DATA_PATH=datasets/mgtv
+export LOGICAL_REASONING_RESULTS_PATH=results/mgtv-results_internlm_best.csv
+
+export USE_FLOAT32_FOR_INFERENCE=false
+
+export USE_BF16_FOR_INFERENCE=false
+echo "Eval $MODEL_NAME with $ADAPTER_NAME_OR_PATH"
+python llm_toolkit/eval_logical_reasoning.py
+
+export USE_BF16_FOR_INFERENCE=true
+echo "Eval $MODEL_NAME with $ADAPTER_NAME_OR_PATH"
+python llm_toolkit/eval_logical_reasoning.py
+
+export LOAD_IN_4BIT=true
+echo "Eval $MODEL_NAME with $ADAPTER_NAME_OR_PATH"
+python llm_toolkit/eval_logical_reasoning.py