Upload folder using huggingface_hub
Browse files- README.md +17 -9
- config.json +5 -3
README.md
CHANGED
|
@@ -1,22 +1,30 @@
|
|
| 1 |
|
| 2 |
-
# MiniGPT
|
| 3 |
|
| 4 |
-
Modelo entrenado
|
| 5 |
|
| 6 |
## Arquitectura
|
| 7 |
|
| 8 |
-
|
| 9 |
-
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
- RoPE
|
| 14 |
- GQA
|
|
|
|
|
|
|
| 15 |
|
| 16 |
## Dataset
|
| 17 |
|
| 18 |
Wikipedia Español
|
| 19 |
|
| 20 |
-
##
|
| 21 |
|
| 22 |
-
|
|
|
|
| 1 |
|
| 2 |
+
# MiniGPT-ES
|
| 3 |
|
| 4 |
+
Modelo causal entrenado en español.
|
| 5 |
|
| 6 |
## Arquitectura
|
| 7 |
|
| 8 |
+
| Campo | Valor |
|
| 9 |
+
|---------|---------|
|
| 10 |
+
| Parámetros | 87.8M |
|
| 11 |
+
| Layers | 12 |
|
| 12 |
+
| Hidden Size | 768 |
|
| 13 |
+
| Attention Heads | 12 |
|
| 14 |
+
| KV Heads | 4 |
|
| 15 |
+
| Context Length | 1024 |
|
| 16 |
+
|
| 17 |
+
## Características
|
| 18 |
+
|
| 19 |
- RoPE
|
| 20 |
- GQA
|
| 21 |
+
- Flash Attention
|
| 22 |
+
- SentencePiece
|
| 23 |
|
| 24 |
## Dataset
|
| 25 |
|
| 26 |
Wikipedia Español
|
| 27 |
|
| 28 |
+
## Framework
|
| 29 |
|
| 30 |
+
PyTorch
|
config.json
CHANGED
|
@@ -1,4 +1,7 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
| 2 |
"model_type": "minigpt",
|
| 3 |
"vocab_size": 16000,
|
| 4 |
"hidden_size": 768,
|
|
@@ -6,7 +9,6 @@
|
|
| 6 |
"num_attention_heads": 12,
|
| 7 |
"num_key_value_heads": 4,
|
| 8 |
"max_position_embeddings": 1024,
|
| 9 |
-
"
|
| 10 |
-
"
|
| 11 |
-
"vocab_file": "tokenizer.model"
|
| 12 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"MiniGPTForCausalLM"
|
| 4 |
+
],
|
| 5 |
"model_type": "minigpt",
|
| 6 |
"vocab_size": 16000,
|
| 7 |
"hidden_size": 768,
|
|
|
|
| 9 |
"num_attention_heads": 12,
|
| 10 |
"num_key_value_heads": 4,
|
| 11 |
"max_position_embeddings": 1024,
|
| 12 |
+
"rope_theta": 10000,
|
| 13 |
+
"torch_dtype": "bfloat16"
|
|
|
|
| 14 |
}
|