NeuralQwen-2.5-1.5B-Spanish is a merge of the following models using
LazyMergekit:
1models:
2 - model: Qwen/Qwen2.5-1.5B
3 # No parameters necessary for base model
4 - model: Qwen/Qwen2.5-1.5B-Instruct
5 parameters:
6 density: 0.66
7 weight: 0.6
8 - model: Kukedlc/Qwen2.5-1.5B-Spanish-1.0-DPO
9 parameters:
10 density: 0.44
11 weight: 0.4
12merge_method: dare_ties
13base_model: Qwen/Qwen2.5-1.5B
14parameters:
15 int8_mask: true
16dtype: float16
1!pip install -qU transformers accelerate
2
3from transformers import AutoTokenizer
4import transformers
5import torch
6
7model = "Kukedlc/NeuralQwen-2.5-1.5B-Spanish"
8messages = [{"role": "system", "content": "Eres un asistente de pensamiento logico que piensa paso a paso, por cada pregunta que te hagan deberes comprobar la respuesta por 3 metodos diferentes."},
9 {"role": "user", "content": "Cuantas letras 'r' tiene la palabra strawberry?"}]
10
11tokenizer = AutoTokenizer.from_pretrained(model)
12prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
13pipeline = transformers.pipeline(
14 "text-generation",
15 model=model,
16 torch_dtype=torch.float16,
17 device_map="auto",
18)
19
20outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
21print(outputs[0]["generated_text"])
