config.json
Browse files- config.json +22 -0
- scripts/pretrain-model.yaml +1 -1
config.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "tangledgroup/tangled-llama-33m-32k-base-v0.1",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"LlamaForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"bos_token_id": 1,
|
| 7 |
+
"eos_token_id": 2,
|
| 8 |
+
"hidden_size": 256,
|
| 9 |
+
"intermediate_size": 1024,
|
| 10 |
+
"max_position_embeddings": 38400,
|
| 11 |
+
"model_type": "llama",
|
| 12 |
+
"num_attention_heads": 32,
|
| 13 |
+
"num_hidden_layers": 4,
|
| 14 |
+
"num_key_value_heads": 8,
|
| 15 |
+
"rms_norm_eps": 1e-05,
|
| 16 |
+
"rope_scaling": null,
|
| 17 |
+
"tie_word_embeddings": false,
|
| 18 |
+
"torch_dtype": "bfloat16",
|
| 19 |
+
"transformers_version": "4.44.2",
|
| 20 |
+
"use_cache": true,
|
| 21 |
+
"vocab_size": 38400
|
| 22 |
+
}
|
scripts/pretrain-model.yaml
CHANGED
|
@@ -78,7 +78,7 @@ train:
|
|
| 78 |
# Total number of tokens to train on (type: Optional[int], default: 3000000000000)
|
| 79 |
# max_tokens: 3000000000000
|
| 80 |
# max_tokens: 8159107755 # 796399 * 2049 * 5
|
| 81 |
-
max_tokens:
|
| 82 |
|
| 83 |
# Limits the number of optimizer steps to run. (type: Optional[int], default: null)
|
| 84 |
max_steps:
|
|
|
|
| 78 |
# Total number of tokens to train on (type: Optional[int], default: 3000000000000)
|
| 79 |
# max_tokens: 3000000000000
|
| 80 |
# max_tokens: 8159107755 # 796399 * 2049 * 5
|
| 81 |
+
max_tokens: 11422750857 # 796399 * 2049 * 7
|
| 82 |
|
| 83 |
# Limits the number of optimizer steps to run. (type: Optional[int], default: null)
|
| 84 |
max_steps:
|