Update README.md

README.md

@@ -1,3 +1,61 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+---
+
+
+```py
+from transformers import AutoConfig, AutoModel, logging
+from transformers import AutoModel, AutoTokenizer
+import torch
+import torch.nn as nn
+from PIL import Image
+import os
+
+logging.set_verbosity_error()  # silence HF info spam
+
+MODEL_ID = "openbmb/MiniCPM-o-2_6"
+
+device = "cpu"
+cfg = AutoConfig.from_pretrained(MODEL_ID, trust_remote_code=True)
+
+cfg.hidden_size = 28 * 6
+cfg.num_heads = 2
+cfg.num_hidden_layers = 2
+cfg.intermediate_size = 32
+
+cfg.vision_config.hidden_size = 32
+cfg.vision_config.num_hidden_layers = 1
+cfg.vision_config.num_attention_heads = 2
+cfg.vision_config.intermediate_size = 32
+cfg.vision_config.image_size = 224
+
+cfg.audio_config.encoder_layers = 2
+
+cfg.tts_config.llm_dim = 32
+cfg.tts_config.hidden_size = 24
+
+model = AutoModel.from_config(cfg, trust_remote_code=True)
+
+print("Built tiny MiniCPM-o model on", device)
+print("Config summary:", {k: getattr(cfg, k) for k in ["hidden_size", "num_hidden_layers", "num_attention_heads", "vocab_size"] if hasattr(cfg, k)})
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
+
+image = Image.open('./image.jpg').convert('RGB')
+question = 'What is in the image?'
+msgs = [{'role': 'user', 'content': [image, question]}]
+
+output_dir = "./tiny-random-minicpmo-new-version"
+os.makedirs(output_dir, exist_ok=True)
+model.save_pretrained(output_dir)
+tokenizer.save_pretrained(output_dir)
+model.processor.save_pretrained(output_dir)
+
+print("Inference starts here")
+res = model.chat(
+    image=None,
+    msgs=msgs,
+    tokenizer=tokenizer
+)
+print(res)
+```