Qwen2.5-3B EN/JA Finetune (Merged)
Full merged checkpoint of a fine-tuned Qwen2.5-3B-Instruct model, trained for English and Japanese use cases.
Model Details
- Base Model:
Qwen/Qwen2.5-3B-Instruct - Architecture:
Qwen2ForCausalLM(36 hidden layers, 16 attention heads, 2 key-value heads) - Parameters: ~3.09B parameters
- Precision:
bfloat16 - Context Length: 32,768 (rope theta 1,000,000)
- Format: Sharded safetensors (
model-00001-of-00002.safetensors,model-00002-of-00002.safetensors)
Quick Usage
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "gradients-io-tournaments/qwen2.5-3b-en-ja-finetune-upwork2"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
device_map="auto"
)
messages = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Hello! / こんにちは！"}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256)
response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
print(response)