jxm commited on
Commit
994b794
·
verified ·
1 Parent(s): 8f96c52

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. config.toml +29 -0
  2. model.pt +3 -0
  3. model_best.pt +3 -0
  4. optimizer.pt +3 -0
config.toml ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ use_wandb = 1
2
+ seed = 1
3
+ style = "dit"
4
+ d_adapter = 256
5
+ normalize_embeddings = 1
6
+ depth = 1
7
+ embs = [ "gte", "gtr", "stella", "sentence-t5", "e5", "sbert", "clip", "stella-big", "snowflake", "ember",]
8
+ n_embs_per_batch = 2
9
+ max_seq_length = 512
10
+ depth_transform = 2
11
+ lr = 5e-5
12
+ bs = 128
13
+ save_every = 400
14
+ epochs = 5.0
15
+ dataset = "nomic_unsupervised"
16
+ max_grad_norm = 10.0
17
+ gradient_accumulation_steps = 1
18
+ loss_coefficient_vsp = 0
19
+ loss_coefficient_contrastive = 1
20
+ loss_coefficient_trans = 1
21
+ loss_coefficient_cc = 0
22
+ eval_steps = 99999999999999999
23
+ cluster_size = 1024
24
+ cluster_strategy = "cluster_within_domain"
25
+ warmup_steps = 100
26
+ wandb_project = "edx-2"
27
+ wandb_name = "dit-pretrain-xs-2"
28
+ save_dir = "checkpoints/{}/"
29
+ num_params = 74724568
model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91b8765486b935bcae4d0662567b52a10ece753cceee17dcfb4ecdc7f2ad6844
3
+ size 299345568
model_best.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ced63a033556e9ea22e6d01a4fb59b7f5e7b92dd8d8d2184273689a547a8ef79
3
+ size 299414082
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3438a42f5e495e5de86325237a56ea6740ed90fda20932b8bfd5a9215151f6a
3
+ size 593418684