WIP

.env.example

@@ -0,0 +1,11 @@
+MODEL_NAME=internlm/internlm2_5-7b-chat-1m
+
+OPENAI_API_KEY=
+HF_TOKEN=
+WANDB_API_KEY=
+
+LOAD_IN_4BIT=false
+NUM_TRAIN_EPOCHS=3
+
+DATA_PATH=datasets/mac
+RESULTS_PATH=results/mac-results.csv

llm_toolkit/eval.py

@@ -14,7 +14,7 @@ path = os.path.dirname(found_dotenv)
 print(f"Adding {path} to sys.path")
 sys.path.append(path)
 
-from llm_toolkit.translation_engine import *
+from llm_toolkit.llm_utils import *
 from llm_toolkit.translation_utils import *
 
 model_name = os.getenv("MODEL_NAME")
@@ -43,6 +43,14 @@ print(f"{start_gpu_memory} GB of memory reserved.")
 
 datasets = load_translation_dataset(data_path, tokenizer)
 
+if len(sys.argv) > 1:
+    num = int(sys.argv[1])
+    if num > 0:
+        print(f"--- evaluating {num} entries")
+        datasets["test"] = datasets["test"].select(range(num))
+
+print_row_details(datasets["test"].to_pandas(), indices=[0, -1])
+
 print("Evaluating model: " + model_name)
 predictions = eval_model(model, tokenizer, datasets["test"])
 

scripts/eval-mac.sh

@@ -0,0 +1,19 @@
+#!/bin/sh
+
+BASEDIR=$(dirname "$0")
+cd $BASEDIR/..
+echo Current Directory:
+pwd
+
+nvidia-smi
+uname -a
+cat /etc/os-release
+lscpu
+grep MemTotal /proc/meminfo
+
+export EVAL_BASE_MODEL=true
+export DO_FINE_TUNING=false
+
+export MODEL_NAME=$1
+echo Evaluating $MODEL_NAME
+python llm_toolkit/tune_mac.py

scripts/install-cuda-torch.sh

@@ -0,0 +1,3 @@
+wget https://developer.download.nvidia.com/compute/cuda/12.4.0/local_installers/cuda_12.4.0_550.54.14_linux.run && \
+sudo sh cuda_12.4.0_550.54.14_linux.run && \
+pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124