Upload folder using huggingface_hub

config.toml

@@ -0,0 +1,66 @@
+seed = 5
+sampling_seed = 5
+train_dataset_seed = 10
+val_dataset_seed = 42
+normalize_embeddings = true
+mixed_precision = "fp16"
+weight_init = "kaiming"
+dataset = "nq"
+max_seq_length = 64
+unsup_emb = "stella"
+sup_emb = "gte"
+n_embs_per_batch = 1
+finetune_mode = false
+noise_level = 0.0
+style = "res_mlp"
+norm_style = "batch"
+depth = 3
+transform_depth = 4
+d_adapter = 1024
+d_hidden = 1024
+d_transform = 1024
+use_small_output_adapters = false
+use_residual_adapters = true
+gan_style = "least_squares"
+disc_depth = 5
+disc_dim = 1024
+use_residual = true
+bs = 256
+gradient_accumulation_steps = 1
+lr = 2e-5
+no_scheduler = true
+max_grad_norm = 1000.0
+loss_coefficient_reverse_rec = 0.0
+loss_coefficient_rec = 1.0
+loss_coefficient_vsp = 1.0
+loss_coefficient_cc_trans = 10.0
+loss_coefficient_cc_rec = 0.0
+loss_coefficient_cc_vsp = 10.0
+loss_coefficient_r1_penalty = 0.0
+warmup_length = 2000
+disc_lr = 1e-5
+eps = 6.25e-10
+smooth = 0.9
+loss_coefficient_disc = 1.0
+loss_coefficient_gen = 1.0
+loss_coefficient_latent_gen = 1.0
+loss_coefficient_similarity_gen = 0.0
+val_size = 4096
+val_bs = 1024
+top_k_size = 1024
+top_k_batches = 4
+k = 16
+heatmap_size = 64
+use_wandb = true
+wandb_project = "unsupervised"
+wandb_name = "n:1000000,e:2000,u:stella,s:gte"
+load_dir = "./finetuning_unsupervised/n:100000,e:10,s:n_double,d:4,d:4,d:64/"
+save_dir = "./finetuning_unsupervised/{}/"
+force_dump = true
+num_points = 1000000
+epochs = 2000
+num_params = 21004800
+num_disc_params = 3943425
+num_sup_disc_params = 3943425
+num_latent_disc_params = 4205569
+num_similarity_disc_params = 3419137

model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3530043f988aeb25b1c9bb9de85ef708fd093827430242f31be3886769d1caa2
+size 84090873