telodigoensergio commited on
Commit
c0311a4
1 Parent(s): 0d944e8

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +45 -2
README.md CHANGED
@@ -87,13 +87,56 @@ Lee el siguiente texto y hazlo más claro:
87
  ###Texto aclarado:
88
  """
89
  ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
90
 
 
91
 
 
 
 
 
 
 
 
 
92
 
93
 
94
  - **Developed by:** [Sergio Chicón](https://huggingface.co/telodigoensergio)
95
- - **Model type:** [More Information Needed]
96
- - **Language(s) (NLP):** Python
97
  - **Finetuned from model:** [Microsoft/phi-2](https://huggingface.co/microsoft/phi-2)
98
 
99
  ### Model Sources
 
87
  ###Texto aclarado:
88
  """
89
  ```
90
+ ## Probar el modelo:
91
+ **Importar las librerías necesarias**:
92
+ ```
93
+ !pip install transformers
94
+ !pip install bitsandbytes
95
+ !pip install accelerate
96
+ import torch
97
+ from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
98
+ ```
99
+
100
+ **Configuración de BitsAndBytes:**
101
+ ```
102
+ bnb_config = BitsAndBytesConfig(load_in_4bit=True,
103
+ bnb_4bit_quant_type='nf4',
104
+ bnb_4bit_compute_dtype='float16',
105
+ bnb_4bit_use_double_quant=True)
106
+ ```
107
+
108
+ **Carga del modelo y el tokenizador:**
109
+ ```
110
+ model = AutoModelForCausalLM.from_pretrained(model_id, device_map='auto',
111
+ quantization_config=bnb_config,
112
+ trust_remote_code=True)
113
+
114
+ tokenizer = AutoTokenizer.from_pretrained(model_id,
115
+ add_eos_token=True,
116
+ trust_remote_code=True)
117
+ tokenizer.pad_token = tokenizer.eos_token
118
+ tokenizer.truncation_side = "left"
119
+ ```
120
+ **Definir prompt e iniciar inferencia:**
121
+ ```
122
+ new_prompt = f"""###System:
123
+ Lee el siguiente texto y hazlo más claro:
124
+ ###Texto:
125
 
126
+ {texto}
127
 
128
+ ###Texto aclarado:
129
+ """
130
+
131
+ inputs = tokenizer(new_prompt, return_tensors="pt", return_attention_mask=False, padding=True, truncation=True)
132
+ outputs = model.generate(**inputs, max_length=1000)
133
+ text = tokenizer.batch_decode(outputs,skip_special_tokens=True)[0]
134
+ print(text)
135
+ ```
136
 
137
 
138
  - **Developed by:** [Sergio Chicón](https://huggingface.co/telodigoensergio)
139
+ - **Language(s) (NLP):** Español(España)
 
140
  - **Finetuned from model:** [Microsoft/phi-2](https://huggingface.co/microsoft/phi-2)
141
 
142
  ### Model Sources