Update README.md

examples/README.md

@@ -14,13 +14,13 @@ Note that you don't need to run all of the scripts in these examples with a CUDA
 but some of the pip packages (like `flash-attn`) will fail to install unless you are installing
 on a device with `nvcc`. Then you can switch to another device. Most of the requirements here are for the plethora of random Hugging Face models from which you might want to get bits-per-byte (BPB) values.
 
-Navigate to `get_error_and_bpb` for examples of getting BPB values and evaluation scores.
+Navigate to `get_error_and_bpb/` for examples of getting BPB values and evaluation scores.
 The examples compute BPB from different groups of ~100 open-source language models on a sample RedPajama V2
 and also a synthetic pretraining dataset. They also compute evals on many different benchmarks using the
 Eleuther Eval Harness.
 
-Navigate to `get_fasttext_filter` for examples of using
-the information from `get_error_and_bpb` to estimate good pretraining
+Navigate to `get_fasttext_filter/` for examples of using
+the information from `get_error_and_bpb/` to estimate good pretraining
 sampling distributions. The examples save reusable fastText training data filters,
 which you can plug into your LLM training pipeline, whatever it may be.