Upload folder using huggingface_hub
Browse files
Llama3_1Base-L3A-32x/checkpoints/final.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:81e142cc2c96956aee630d8f8cd6d75629ec6a71fb9bdffe0ef2c2652fd1effe
|
3 |
+
size 2148278848
|
Llama3_1Base-L3A-32x/hyperparams.json
ADDED
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"hook_point_in": "blocks.3.hook_attn_out",
|
3 |
+
"hook_point_out": "blocks.3.hook_attn_out",
|
4 |
+
"use_decoder_bias": true,
|
5 |
+
"apply_decoder_bias_to_pre_encoder": false,
|
6 |
+
"expansion_factor": 32,
|
7 |
+
"d_model": 4096,
|
8 |
+
"d_sae": 131072,
|
9 |
+
"norm_activation": "token-wise",
|
10 |
+
"dataset_average_activation_norm": null,
|
11 |
+
"decoder_exactly_fixed_norm": false,
|
12 |
+
"sparsity_include_decoder_norm": true,
|
13 |
+
"use_glu_encoder": false,
|
14 |
+
"init_decoder_norm": null,
|
15 |
+
"init_encoder_norm": null,
|
16 |
+
"init_encoder_with_decoder_transpose": true,
|
17 |
+
"l1_coefficient": 1.6e-05,
|
18 |
+
"l1_coefficient_warmup_steps": 29296,
|
19 |
+
"lp": 1,
|
20 |
+
"use_ghost_grads": false,
|
21 |
+
"tp_size": 1,
|
22 |
+
"ddp_size": 1
|
23 |
+
}
|
Llama3_1Base-L3A-32x/lm_config.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "meta-llama/Meta-Llama-3.1-8B",
|
3 |
+
"use_flash_attn": false,
|
4 |
+
"cache_dir": null,
|
5 |
+
"d_model": 4096,
|
6 |
+
"local_files_only": false
|
7 |
+
}
|