run formatter

README.md

@@ -65,7 +65,7 @@ Both scripts share the same arguments (except for `--fp8` quantization, which on
 ### Example 1: Single prompt (use transformers)
 ```bash
 python generate_transformers.py \
-    --prompt "A cheerful piano melody"
+    --prompt "A cheerful rock song with bright electric guitars"
 ```
 Outputs 4 MIDIs (and synthesized MP3s) conditioned on the same prompt by default.
 

assets/example_prompts.txt

@@ -1,5 +0,0 @@
-Rock piece suitable for a hero's epic battle against a dark villain. The music is rich with synth leads, brass section, and heavy drums.
-A sad and emotional song featuring relentless, heartbreaking acoustic guitars.
-Slow and loving music that is best for a romantic film. The piano and string ensemble give a classical vibe.
-An energetic and motivating pop song you love to hear on a long road trip.
-Upbeat and playful jazz music with lively saxophones, like you're going out on a Sunday picnic.

generate_transformers.py

@@ -339,7 +339,7 @@ Examples:
     print(f"Output directory: {output_dir.absolute()}\n")
 
     # Load tokenizer from the model checkpoint
-    print("Loading tokenizer...")
+    print("Loading tokenizer and model...")
     tokenizer = AutoTokenizer.from_pretrained(
         args.model,
         cache_dir=args.cache_dir,
@@ -471,4 +471,3 @@ Examples:
 
 if __name__ == "__main__":
     main()
-

generate_vllm.py

@@ -366,7 +366,7 @@ Examples:
     print(f"Output directory: {output_dir.absolute()}\n")
 
     # Load tokenizer from the model checkpoint (not base Llama!)
-    print("Loading tokenizer...")
+    print("Loading tokenizer and model...")
     tokenizer = AutoTokenizer.from_pretrained(
         args.model,  # Use the model checkpoint path, not base Llama
         cache_dir=args.cache_dir,