README.md

metadata

language:
  - th
  - en
library_name: transformers
base_model:
  - Qwen/Qwen2.5-7B-Instruct
  - Qwen/Qwen2.5-7B
pipeline_tag: text-generation
model-index:
  - name: Tsunami-0.5-7B-Instruct
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: IFEval (0-Shot)
          type: HuggingFaceH4/ifeval
          args:
            num_few_shot: 0
        metrics:
          - type: inst_level_strict_acc and prompt_level_strict_acc
            value: 74
            name: strict accuracy
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=Tsunami-th/Tsunami-0.5-7B-Instruct
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: BBH (3-Shot)
          type: BBH
          args:
            num_few_shot: 3
        metrics:
          - type: acc_norm
            value: 36.14
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=Tsunami-th/Tsunami-0.5-7B-Instruct
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MATH Lvl 5 (4-Shot)
          type: hendrycks/competition_math
          args:
            num_few_shot: 4
        metrics:
          - type: exact_match
            value: 0.15
            name: exact match
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=Tsunami-th/Tsunami-0.5-7B-Instruct
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GPQA (0-shot)
          type: Idavidrein/gpqa
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 7.83
            name: acc_norm
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=Tsunami-th/Tsunami-0.5-7B-Instruct
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MuSR (0-shot)
          type: TAUR-Lab/MuSR
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 12.21
            name: acc_norm
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=Tsunami-th/Tsunami-0.5-7B-Instruct
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU-PRO (5-shot)
          type: TIGER-Lab/MMLU-Pro
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 37.92
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=Tsunami-th/Tsunami-0.5-7B-Instruct
          name: Open LLM Leaderboard
license: apache-2.0

Tsunami-0.5-7B-Instruct

TSUNAMI: Transformative Semantic Understanding and Natural Augmentation Model for Intelligence.

TSUNAMI full name was created by ChatGPT.

---

infomation

Tsunami-0.5-7B-Instruct is Thai Large Language Model that fine-tuned from Qwen2.5-7B around 60,000 rows in Thai-specific domain.

---

Prompt Template

This model uses ChatML prompt template:

<|im_start|>system
{System}<|im_end|>
<|im_start|>user
{User}<|im_end|>
<|im_start|>assistant
{Assistant}

How to use

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "Tsunami-th/Tsunami-0.5-7B-Instruct"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "สวัสดีครับ"}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

inputs = tokenizer(text, return_tensors="pt")
inputs = inputs.to(model.device)
with torch.no_grad():
   output = model.generate(**inputs, max_new_tokens=512)

response = tokenizer.decode(output[0, len(inputs['input_ids'][0]):], skip_special_tokens=True)

---

Author

• Pollakrit Lorprasertkul | [email protected]

---

• Tsunami-0.5-7B-Instruct is the version 0.5 that did not train on the whole dataset.
• Tsunami-1.0-7B-Instruct is coming soon.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	28.04
IFEval (0-Shot)	74.00
BBH (3-Shot)	36.14
MATH Lvl 5 (4-Shot)	0.15
GPQA (0-shot)	7.83
MuSR (0-shot)	12.21
MMLU-PRO (5-shot)	37.92