myamafuj/gemma-2-27b-it

このモデルを用いた出力例

※Omnicampusの環境では依存環境が壊れてunslothが使えなかったのでColab L4での再現方法

!pip install --no-cache-dir unsloth==2024.12.4 ipywidgets flash-attn==2.7.0.post2

import json

from tqdm import tqdm
from unsloth import FastLanguageModel


model_name = 'myamafuj/gemma-2-27b-it'
max_seq_length = 2048
dtype = None
load_in_4bit = True

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = model_name,
    max_seq_length = max_seq_length,
    dtype = dtype,
    load_in_4bit = load_in_4bit,
    trust_remote_code=True,
)
FastLanguageModel.for_inference(model)

datasets = []
with open("/content/elyza-tasks-100-TV_0.jsonl", "r") as f:
    item = ""
    for line in f:
      line = line.strip()
      item += line
      if item.endswith("}"):
        datasets.append(json.loads(item))
        item = ""

prompt_template = \
"""###以下は、タスクを説明する指示です。要求を適切に満たす応答を書きなさい。

### 指示:
{}

### 応答:
"""
results = []
for dt in tqdm(datasets):
  input = dt["input"]

  prompt = prompt_template.format(input)

  inputs = tokenizer([prompt], return_tensors="pt").to(model.device)

  outputs = model.generate(**inputs, max_new_tokens=512, use_cache=True, do_sample=False, repetition_penalty=1.2)
  prediction = tokenizer.decode(outputs[0], skip_special_tokens=True).split('\n### 応答:')[-1]

  results.append({"task_id": dt["task_id"], "input": input, "output": prediction})

with open(f"{model_name}_output.jsonl".split('/')[-1], 'w', encoding='utf-8') as f:
    for result in results:
        json.dump(result, f, ensure_ascii=False)
        f.write('\n')

---

base_model: unsloth/gemma-2-27b-bnb-4bit tags: - text-generation-inference - transformers - unsloth - gemma2 - trl license: apache-2.0 language: - en

Uploaded model

• Developed by: myamafuj
• License: apache-2.0
• Finetuned from model : unsloth/gemma-2-27b-bnb-4bit

This gemma2 model was trained 2x faster with Unsloth and Huggingface's TRL library.