add sample and test

README.md

@@ -8,5 +8,14 @@ ref: https://huggingface.co/elyza/ELYZA-japanese-Llama-2-13b-fast-instruct/blob/
 量子化の詳細はconfig.jsonを参照してください。  
 コードはquantize.pyです。  
 
+
+## Test
+elyza/ELYZA-tasks-100 の実行結果
+→ task100.csv
+
+実行時間
+開始: 2024-05-09 03:49:30.779783
+終了: 2024-05-09 03:56:56.380691
+
 ## LICENSE
 LICENSE: LLAMA2  

sample.py

@@ -0,0 +1,34 @@
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+B_INST, E_INST = "[INST]", "[/INST]"
+B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"
+DEFAULT_SYSTEM_PROMPT = "あなたは誠実で優秀な日本人のアシスタントです。"
+text = "仕事の熱意を取り戻すためのアイデアを5つ挙げてください。"
+
+model_name = "./"
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    use_cache=True,
+    device_map="auto"
+)
+model.eval()
+
+prompt = "{bos_token}{b_inst} {system}{prompt} {e_inst} ".format(
+    bos_token=tokenizer.bos_token,
+    b_inst=B_INST,
+    system=f"{B_SYS}{DEFAULT_SYSTEM_PROMPT}{E_SYS}",
+    prompt=text,
+    e_inst=E_INST,
+)
+token_ids = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt")
+output_ids = model.generate(
+    token_ids.to(model.device),
+    max_new_tokens=256,
+    pad_token_id=tokenizer.pad_token_id,
+    eos_token_id=tokenizer.eos_token_id,
+)
+output = tokenizer.decode(output_ids.tolist()[0][token_ids.size(1) :], skip_special_tokens=True)
+print(output)

task100.py

@@ -0,0 +1,55 @@
+import torch
+from datasets import load_dataset
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import datetime
+
+B_INST, E_INST = "[INST]", "[/INST]"
+B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"
+DEFAULT_SYSTEM_PROMPT = "あなたは誠実で優秀な日本人のアシスタントです。"
+
+
+def main():
+    model_name = "./"
+    ds = load_dataset("elyza/ELYZA-tasks-100", revision="1.0.0")
+
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    model = AutoModelForCausalLM.from_pretrained(
+        model_name,
+        use_cache=True,
+        device_map="auto"
+    )
+    model.eval()
+
+    def pred(example):
+        prompt = "{bos_token}{b_inst} {system}{prompt}{e_inst} ".format(
+            bos_token=tokenizer.bos_token,
+            b_inst=B_INST,
+            system=f"{B_SYS}{DEFAULT_SYSTEM_PROMPT}{E_SYS}",
+            prompt=example["input"],
+            e_inst=E_INST,
+        )
+
+        token_ids = tokenizer.encode(
+            prompt, add_special_tokens=False, return_tensors="pt"
+        )
+
+        with torch.no_grad():
+            output_ids = model.generate(
+                token_ids.to(model.device),
+                max_new_tokens=1200,
+                pad_token_id=tokenizer.pad_token_id,
+                eos_token_id=tokenizer.eos_token_id,
+            )
+        output = tokenizer.decode(
+            output_ids.tolist()[0][token_ids.size(1) :], skip_special_tokens=True
+        )
+        example[model_name] = output
+        return example
+    print(datetime.datetime.now())
+    ds = ds.map(pred, batched=False)
+    print(datetime.datetime.now())
+    ds["test"].to_csv(f"preds/{model_name.replace('/', '-')}.csv", index=False)
+
+
+if __name__ == "__main__":
+    main()