Update README.md

README.md

@@ -8,6 +8,24 @@
 from transformers.models.qwen3_omni_moe import Qwen3OmniMoeThinkerTextModel
 from transformers.models.qwen3_omni_moe.configuration_qwen3_omni_moe import Qwen3OmniMoeTextConfig 
 # 1. init orig model
+import os
+os.environ["CUDA_VISIBLE_DEVICES"]='0'
+import soundfile as sf
+
+from transformers import Qwen3OmniMoeForConditionalGeneration, Qwen3OmniMoeProcessor
+from qwen_omni_utils import process_mm_info
+
+MODEL_PATH = "/scratch/vladimir_albrekht/qwen3omni-fine-tuning/models/Qwen3-Omni-30B-A3B-Instruct"
+# MODEL_PATH = "Qwen/Qwen3-Omni-30B-A3B-Thinking"
+
+model = Qwen3OmniMoeForConditionalGeneration.from_pretrained(
+    MODEL_PATH,
+    dtype="auto",
+    device_map="auto",
+    attn_implementation="flash_attention_2",
+)
+
+processor = Qwen3OmniMoeProcessor.from_pretrained(MODEL_PATH)
 
 # 2. Get the original config and create a new, smaller version
 # where model == qwen3omni orig 30B model  `Qwen/Qwen3-Omni-30B-A3B-Instruct`
@@ -45,5 +63,6 @@ model.config.thinker_config.text_config = small_thinker_llm.config
 PATH = "./qwen3omni_5B_random_thinker_weights"
 model.save_pretrained(PATH)
 processor.save_pretrained(PATH)
+
 ```