Delete AmInSide

AmInSide/0_data_gen/app.py

@@ -1,14 +0,0 @@
-from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
-import gradio as gr
-
-model_id = "Hodely/AmInSide1.0"  # Reemplaza con tu repo real
-
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForCausalLM.from_pretrained(model_id)
-pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
-
-def chat(user_input):
-    result = pipe(user_input, max_new_tokens=200, temperature=0.7, do_sample=True)
-    return result[0]['generated_text']
-
-gr.Interface(fn=chat, inputs="text", outputs="text", title="🧠 AmInSide1.0").launch()

AmInSide/0_data_gen/generate_dataset.py

@@ -1,22 +0,0 @@
-from transformers import pipeline
-import json
-
-generator = pipeline("text-generation", model="mistralai/Mistral-7B-Instruct", tokenizer="mistralai/Mistral-7B-Instruct")
-
-seed_prompts = [
-    "Dame 10 preguntas filosóficas con respuestas profundas.",
-    "Genera 5 ejemplos tipo ChatGPT con tono sarcástico pero sabio.",
-    "Crea 10 instrucciones para IA educativa con respuestas creativas."
-]
-
-output = []
-for prompt in seed_prompts:
-    result = generator(prompt, max_new_tokens=512)[0]["generated_text"]
-    # ¡Aquí puedes separar y limpiar! Por ahora simplificamos
-    output.append({"instruction": prompt, "response": result})
-
-with open("instruct_dataset.jsonl", "w", encoding="utf-8") as f:
-    for example in output:
-        f.write(json.dumps(example, ensure_ascii=False) + "\n")
-
-print("✅ Dataset generado.")

AmInSide/1_train/train.py

@@ -1,35 +0,0 @@
-from datasets import load_dataset
-from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, Trainer
-from peft import get_peft_model, LoraConfig, TaskType
-import torch
-
-model_id = "microsoft/phi-3-mini-4k-instruct"
-dataset_path = "../0_data_gen/instruct_dataset.jsonl"
-
-# Carga dataset personalizado
-data = load_dataset("json", data_files=dataset_path)
-
-# Tokenización
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-def tokenize(example):
-    return tokenizer(f"<|user|>{example['instruction']}<|assistant|>{example['response']}", truncation=True, padding="max_length", max_length=512)
-tokenized = data["train"].map(tokenize)
-
-# Carga modelo + PEFT
-model = AutoModelForCausalLM.from_pretrained(model_id)
-peft_config = LoraConfig(task_type=TaskType.CAUSAL_LM, inference_mode=False, r=8, lora_alpha=16, lora_dropout=0.05)
-model = get_peft_model(model, peft_config)
-
-# Entrenamiento
-training_args = TrainingArguments(
-    output_dir="./model",
-    per_device_train_batch_size=2,
-    num_train_epochs=3,
-    save_total_limit=1,
-    logging_steps=10,
-    learning_rate=2e-4,
-    fp16=torch.cuda.is_available()
-)
-
-trainer = Trainer(model=model, args=training_args, train_dataset=tokenized)
-trainer.train()

AmInSide/2_space/README.md

@@ -1,6 +0,0 @@
-# AmInSide1.0
-Este es el modelo de lenguaje creado por Hodely. Entrenado con ejemplos generados por otro modelo, afinado para combinar inteligencia, humor y creatividad.
-
-🔹 Entrenado con LoRA  
-🔹 Base: Phi-3 / Mistral  
-🔹 Fine-tuned en estilo instructivo con dataset propio

AmInSide/2_space/app.py

@@ -1,14 +0,0 @@
-from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
-import gradio as gr
-
-model_id = "Hodely/AmInSide1.0"  # Reemplaza con tu repo real
-
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForCausalLM.from_pretrained(model_id)
-pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
-
-def chat(user_input):
-    result = pipe(user_input, max_new_tokens=200, temperature=0.7, do_sample=True)
-    return result[0]['generated_text']
-
-gr.Interface(fn=chat, inputs="text", outputs="text", title="🧠 AmInSide1.0").launch()

AmInSide/2_space/requirements.txt

@@ -1,5 +0,0 @@
-transformers
-gradio
-peft
-datasets
-accelerate