Auto Commit
Browse files- config.json +55 -0
- model_final.pth +3 -0
config.json
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"n_layers": 2,
|
| 3 |
+
"d_model": 256,
|
| 4 |
+
"d_mlp": 1024,
|
| 5 |
+
"d_head": 64,
|
| 6 |
+
"n_heads": 4,
|
| 7 |
+
"lr_hidden": 0.002,
|
| 8 |
+
"lr_vector": 0.001,
|
| 9 |
+
"batch_size_per_device": 32,
|
| 10 |
+
"batches_per_step": 1,
|
| 11 |
+
"seed": 9023,
|
| 12 |
+
"save_checkpoints": true,
|
| 13 |
+
"debug": false,
|
| 14 |
+
"debug_batch": false,
|
| 15 |
+
"normalization": "LN",
|
| 16 |
+
"max_tokens": 22000000000,
|
| 17 |
+
"version": 178,
|
| 18 |
+
"use_bfloat16_matmul": true,
|
| 19 |
+
"n_ctx": 1024,
|
| 20 |
+
"d_vocab": 50278,
|
| 21 |
+
"tokenizer_name": "EleutherAI/gpt-neox-20b",
|
| 22 |
+
"betas": [
|
| 23 |
+
0.9,
|
| 24 |
+
0.99
|
| 25 |
+
],
|
| 26 |
+
"weight_decay": 0.05,
|
| 27 |
+
"dataset_name": "the_pile",
|
| 28 |
+
"grad_norm_clip": 1.0,
|
| 29 |
+
"n_devices": 8,
|
| 30 |
+
"act_fn": "solu_ln",
|
| 31 |
+
"shortformer_pos": true,
|
| 32 |
+
"attn_only": true,
|
| 33 |
+
"ln_eps": 1e-05,
|
| 34 |
+
"lr_schedule": "cosine_warmup",
|
| 35 |
+
"warmup_tokens": 300000000,
|
| 36 |
+
"train_loss_ewma_beta": 0.99,
|
| 37 |
+
"truncate_tokens": 5000000,
|
| 38 |
+
"log_interval": 50,
|
| 39 |
+
"initializer_scale_global": 1.0,
|
| 40 |
+
"initializer_scale_hidden": 0.02,
|
| 41 |
+
"initializer_scale_embed": 0.1,
|
| 42 |
+
"initializer_scale_unembed": 0.02,
|
| 43 |
+
"neuron_scale": 1.0,
|
| 44 |
+
"neuron_temp": 1.0,
|
| 45 |
+
"use_acc": false,
|
| 46 |
+
"weight_init_scheme": "gpt2",
|
| 47 |
+
"fixed_init": "",
|
| 48 |
+
"store_init": true,
|
| 49 |
+
"control": 1.0,
|
| 50 |
+
"tokens_per_step": 262144,
|
| 51 |
+
"batch_size": 256,
|
| 52 |
+
"max_steps": 83923,
|
| 53 |
+
"warmup_steps": 1144,
|
| 54 |
+
"n_params": 393216
|
| 55 |
+
}
|
model_final.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e02fc02129aa2d471f20e260aea788c23be02fb742259dbdc85001404d03b9dc
|
| 3 |
+
size 108427549
|