Upload config.json with huggingface_hub
Browse files- config.json +24 -0
config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architecture": "Raschka GPTModel (separate W_query/W_key/W_value, no weight tying)",
|
| 3 |
+
"model_type": "instruction-tuned (SFT)",
|
| 4 |
+
"base_model": "nishantup/nanogpt-slm-124m (gpt_slm_best.pth)",
|
| 5 |
+
"model_config": {
|
| 6 |
+
"vocab_size": 50257,
|
| 7 |
+
"context_length": 256,
|
| 8 |
+
"emb_dim": 768,
|
| 9 |
+
"n_heads": 12,
|
| 10 |
+
"n_layers": 12,
|
| 11 |
+
"drop_rate": 0.0,
|
| 12 |
+
"qkv_bias": false
|
| 13 |
+
},
|
| 14 |
+
"total_parameters_millions": 163.2,
|
| 15 |
+
"tokenizer": "tiktoken gpt2 (50,257 BPE tokens)",
|
| 16 |
+
"framework": "PyTorch",
|
| 17 |
+
"prompt_format": "Alpaca (### Instruction / ### Input / ### Response)",
|
| 18 |
+
"training": {
|
| 19 |
+
"dataset": "Alpaca-format instruction dataset (1,100 examples)",
|
| 20 |
+
"epochs": 2,
|
| 21 |
+
"optimizer": "AdamW (lr=5e-5, weight_decay=0.1)",
|
| 22 |
+
"max_length": 256
|
| 23 |
+
}
|
| 24 |
+
}
|