Update usage

README.md

@@ -52,7 +52,7 @@ from mantis.models.mllava import chat_mllava, LlavaForConditionalGeneration, MLl
 from mantis.models.mllava.utils import conv_templates
 from transformers import AutoTokenizer
 
-# Set the system prompt
+# 1. Set the system prompt
 conv_llama_3_elyza = Conversation(
     system="<|start_header_id|>system<|end_header_id|>\n\nあなたは誠実で優秀な日本人のアシスタントです。特に指示が無い場合は、常に日本語で回答してください。",
     roles=("user", "assistant"),
@@ -63,7 +63,7 @@ conv_llama_3_elyza = Conversation(
 )
 conv_templates["llama_3"] = conv_llama_3_elyza
 
-# 1. load model
+# 2. Load model
 device = "cuda" if torch.cuda.is_available() else "cpu"
 model_id = "SakanaAI/Llama-3-EvoVLM-JP-v2"
 
@@ -72,17 +72,7 @@ processor.tokenizer.pad_token = processor.tokenizer.eos_token
 
 model = LlavaForConditionalGeneration.from_pretrained(model_id, torch_dtype=torch.float16, device_map=device).eval()
 
-# 2. prepare inputs
-text = "<image> <image>それぞれ簡単に説明してください。"
-url_list = [
-    "https://images.unsplash.com/photo-1703023190744-a28366dfcebe?q=80&w=2574&auto=format&fit=crop&ixlib=rb-4.0.3&ixid=M3wxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8fA%3D%3D",
-    "https://images.unsplash.com/photo-1541364983171-a8ba01e95cfc?q=80&w=2574&auto=format&fit=crop&ixlib=rb-4.0.3&ixid=M3wxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8fA%3D%3D"
-]
-images = [
-   Image.open(requests.get(url, stream=True).raw).convert("RGB") for url in url_list
-]
-
-# 3. generate
+# 3. Prepare a generate config
 generation_kwargs = {
     "max_new_tokens": 128,
     "num_beams": 1,
@@ -90,10 +80,30 @@ generation_kwargs = {
     "no_repeat_ngram_size": 3,
 }
 
+# 4. Generate
+text = "<image>の信号は何色ですか？"
+url_list = [
+    "https://images.unsplash.com/photo-1694831404826-3400c48c188d?q=80&w=2070&auto=format&fit=crop&ixlib=rb-4.0.3&ixid=M3wxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8fA%3D%3D",
+    "https://images.unsplash.com/photo-1693240876439-473af88b4ed7?q=80&w=1974&auto=format&fit=crop&ixlib=rb-4.0.3&ixid=M3wxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8fA%3D%3D"
+]
+images = [
+   Image.open(requests.get(url_list[0], stream=True).raw).convert("RGB")
+]
+
 response, history = chat_mllava(text, images, model, processor, **generation_kwargs)
 
 print(response)
-# 1は、クリスマス帽子を被ったフランダースブルドッグの写真です。2は、眼鏡を掛けたフランケンブルドックの写真で、どちらも可愛くて面白いです。
+# 信号の色は、青色です。
+
+# 5. Multi-turn conversation
+text = "では、<image>の信号は？"
+images += [
+   Image.open(requests.get(url_list[1], stream=True).raw).convert("RGB")
+]
+response, history = chat_mllava(text, images, model, processor, history=history, **generation_kwargs)
+
+print(response)
+# 赤色
 ```
 
 </details>