Fix evaluation.md checkpoint dirs

tutorials/evaluation.md

@@ -20,7 +20,7 @@ Use the following command to evaluate Lit-GPT models on all tasks in Eleuther AI
 
 ```bash
 python eval/lm_eval_harness.py \
-    --checkpoint_dir "checkpoints/Llama-2-7b-hf/" \
+    --checkpoint_dir "checkpoints/meta-llama/Llama-2-7b-hf" \
     --precision "bf16-true" \
     --batch_size 4 \
     --save_filepath "results.json"
@@ -30,7 +30,7 @@ To evaluate on LLMs on specific tasks, for example, TruthfulQA and HellaSwag, yo
 
 ```bash
 python eval/lm_eval_harness.py \
-    --checkpoint_dir "checkpoints/Llama-2-7b-hf/" \
+    --checkpoint_dir "checkpoints/meta-llama/Llama-2-7b-hf" \
     --eval_tasks "[truthfulqa_mc,hellaswag]" \
     --precision "bf16-true" \
     --batch_size 4 \
@@ -57,7 +57,7 @@ For LoRA-finetuned models, you need to first merge the LoRA weights with the ori
 
   ```shell
   python eval/lm_eval_harness.py \
-      --checkpoint_dir "checkpoints/Llama-2-7b-hf/" \
+      --checkpoint_dir "checkpoints/meta-llama/Llama-2-7b-hf" \
       --precision "bf16-true" \
       --eval_tasks "[hendrycksTest*]" \
       --batch_size 4 \