Commit
•
b746b12
1
Parent(s):
12d65a0
Upload jax_model/hyperparams.json with huggingface_hub
Browse files- jax_model/hyperparams.json +43 -0
jax_model/hyperparams.json
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset_name": "multi_species_final_178G.upper.n.filtered.6_mers.1000_tok_len.overlap_50",
|
| 3 |
+
"alphabet": "k-mers",
|
| 4 |
+
"k_for_kmers": 6,
|
| 5 |
+
"num_warmup_updates": 16000,
|
| 6 |
+
"warmup_init_lr": 5e-05,
|
| 7 |
+
"warmup_end_lr": 0.0001,
|
| 8 |
+
"training_set_proportion": 0.95,
|
| 9 |
+
"tokens_per_batch": 1000000.0,
|
| 10 |
+
"tokens_per_checkpoint": "3500000000.0",
|
| 11 |
+
"masking_ratio": 0.15,
|
| 12 |
+
"masking_prob": 0.8,
|
| 13 |
+
"batch_size": 2,
|
| 14 |
+
"random_token_prob": 0.1,
|
| 15 |
+
"dropout_rate": 0.1,
|
| 16 |
+
"num_hosts": 8,
|
| 17 |
+
"server_address": "3422826-worker-0:1234",
|
| 18 |
+
"alphabet_size": 4105,
|
| 19 |
+
"pad_token_id": 1,
|
| 20 |
+
"mask_token_id": 2,
|
| 21 |
+
"class_token_id": 3,
|
| 22 |
+
"eos_token_id": -1,
|
| 23 |
+
"prepend_bos": true,
|
| 24 |
+
"append_eos": false,
|
| 25 |
+
"max_positions": 1000,
|
| 26 |
+
"emb_layer_norm_before": false,
|
| 27 |
+
"attention_heads": 20,
|
| 28 |
+
"embed_dim": 2560,
|
| 29 |
+
"ffn_embed_dim": 10240,
|
| 30 |
+
"num_layers": 32,
|
| 31 |
+
"token_dropout": true,
|
| 32 |
+
"embed_scale": 1.0,
|
| 33 |
+
"use_remat": false,
|
| 34 |
+
"architecture": "Vanilla",
|
| 35 |
+
"acc_batch_size": 8,
|
| 36 |
+
"num_local_devices": 8,
|
| 37 |
+
"num_global_devices": 64,
|
| 38 |
+
"tokens_length": 1000,
|
| 39 |
+
"mixed-precision": true,
|
| 40 |
+
"model_num_parameters": "2547800585",
|
| 41 |
+
"shift": true,
|
| 42 |
+
"overlap": 50
|
| 43 |
+
}
|