Update README.md
Browse files
README.md
CHANGED
|
@@ -17,7 +17,7 @@ tags:
|
|
| 17 |
language:
|
| 18 |
- es
|
| 19 |
datasets:
|
| 20 |
-
-
|
| 21 |
base_model: mistralai/Mistral-Small-3.2-24B-Instruct-2506
|
| 22 |
inference: true
|
| 23 |
---
|
|
@@ -32,7 +32,7 @@ KAL-24B-mx-v1 es un Large Language Model (LLM) basado en la arquitectura Mistral
|
|
| 32 |
|
| 33 |
**Modelo Base:** Mistral-Small-3.2-24B-Instruct-2506
|
| 34 |
|
| 35 |
-
**Identificador en Hugging Face:**
|
| 36 |
|
| 37 |
**Licencia:** Apache 2.0
|
| 38 |
|
|
@@ -116,7 +116,7 @@ Este modelo no está diseñado para:
|
|
| 116 |
|
| 117 |
### Fuentes del Modelo
|
| 118 |
|
| 119 |
-
- **Repositorio:**
|
| 120 |
|
| 121 |
- **Modelo Base:** mistralai/Mistral-Small-3.2-24B-Instruct-2506
|
| 122 |
|
|
@@ -218,7 +218,7 @@ pip install "transformers>=4.57.1" "torch>=2.2" "accelerate>=1.0" "mistral-commo
|
|
| 218 |
from transformers import AutoTokenizer, Mistral3ForConditionalGeneration
|
| 219 |
import torch
|
| 220 |
|
| 221 |
-
MODEL_ID = "
|
| 222 |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
|
| 223 |
if tokenizer.pad_token is None:
|
| 224 |
tokenizer.pad_token = tokenizer.eos_token
|
|
@@ -269,7 +269,7 @@ print(chat("¿Cómo puedo obtener mi CURP por primera vez?"))
|
|
| 269 |
```bash
|
| 270 |
pip install -U "vllm>=0.9.1"
|
| 271 |
|
| 272 |
-
vllm serve
|
| 273 |
--dtype bfloat16 \
|
| 274 |
--tokenizer-mode mistral \
|
| 275 |
--config-format mistral \
|
|
@@ -288,7 +288,7 @@ from openai import OpenAI
|
|
| 288 |
client = OpenAI(api_key="EMPTY", base_url="http://localhost:8000/v1")
|
| 289 |
|
| 290 |
response = client.chat.completions.create(
|
| 291 |
-
model="
|
| 292 |
messages=[
|
| 293 |
{"role": "system", "content": ""Eres KAL, la IA más inteligente de México. Responde con precisión y brevedad."},
|
| 294 |
{"role": "user", "content": "¿Qué necesito para tramitar mi RFC?"}
|
|
@@ -498,6 +498,6 @@ Si usas este modelo en tu trabajo, por favor cítalo de la siguiente manera:
|
|
| 498 |
author={Saptiva AI},
|
| 499 |
year={2025},
|
| 500 |
publisher={Hugging Face},
|
| 501 |
-
howpublished={\url{https://huggingface.co/
|
| 502 |
}
|
| 503 |
```
|
|
|
|
| 17 |
language:
|
| 18 |
- es
|
| 19 |
datasets:
|
| 20 |
+
- SaptivaAI/kal-mx-training-data
|
| 21 |
base_model: mistralai/Mistral-Small-3.2-24B-Instruct-2506
|
| 22 |
inference: true
|
| 23 |
---
|
|
|
|
| 32 |
|
| 33 |
**Modelo Base:** Mistral-Small-3.2-24B-Instruct-2506
|
| 34 |
|
| 35 |
+
**Identificador en Hugging Face:** SaptivaAI/KAL-24B-mx-v1
|
| 36 |
|
| 37 |
**Licencia:** Apache 2.0
|
| 38 |
|
|
|
|
| 116 |
|
| 117 |
### Fuentes del Modelo
|
| 118 |
|
| 119 |
+
- **Repositorio:** SaptivaAI/KAL-24B-mx-v1
|
| 120 |
|
| 121 |
- **Modelo Base:** mistralai/Mistral-Small-3.2-24B-Instruct-2506
|
| 122 |
|
|
|
|
| 218 |
from transformers import AutoTokenizer, Mistral3ForConditionalGeneration
|
| 219 |
import torch
|
| 220 |
|
| 221 |
+
MODEL_ID = "SaptivaAI/KAL-24B-mx-v1"
|
| 222 |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
|
| 223 |
if tokenizer.pad_token is None:
|
| 224 |
tokenizer.pad_token = tokenizer.eos_token
|
|
|
|
| 269 |
```bash
|
| 270 |
pip install -U "vllm>=0.9.1"
|
| 271 |
|
| 272 |
+
vllm serve SaptivaAI/KAL-24B-mx-v1 \
|
| 273 |
--dtype bfloat16 \
|
| 274 |
--tokenizer-mode mistral \
|
| 275 |
--config-format mistral \
|
|
|
|
| 288 |
client = OpenAI(api_key="EMPTY", base_url="http://localhost:8000/v1")
|
| 289 |
|
| 290 |
response = client.chat.completions.create(
|
| 291 |
+
model="SaptivaAI/KAL-24B-mx-v1",
|
| 292 |
messages=[
|
| 293 |
{"role": "system", "content": ""Eres KAL, la IA más inteligente de México. Responde con precisión y brevedad."},
|
| 294 |
{"role": "user", "content": "¿Qué necesito para tramitar mi RFC?"}
|
|
|
|
| 498 |
author={Saptiva AI},
|
| 499 |
year={2025},
|
| 500 |
publisher={Hugging Face},
|
| 501 |
+
howpublished={\url{https://huggingface.co/SaptivaAI/KAL-24B-mx-v1}}
|
| 502 |
}
|
| 503 |
```
|