Upload folder using huggingface_hub

klbxe18a/conf.yaml

@@ -0,0 +1,51 @@
+data:
+  batch_size: 6
+  num_workers: 2
+  per_gpu_batch_size: 4
+deploy: true
+device: cuda:0
+device_ids:
+- 0
+epochs: 1
+eval:
+  save_tables: false
+log:
+  eval_interval: 50000
+  log_interval: 10
+  save_interval: 50000
+  save_multiple: false
+num_gpus: 1
+optimizer:
+  beta1: 0.9
+  beta2: 0.95
+  decay_lr: true
+  grad_clip: 1.0
+  learning_rate: 0.001
+  min_lr: 0.0009
+  warmup_iters: 200
+  weight_decay: 0.0001
+precomp:
+  activation_scaling_factor: 0.05914954780170705
+  cache_dir: /home/can/dictionary_learning/precomputed_activations_gemma_l12_pile_100_000_000_tokens
+  context_length: 500
+  do_distributed_training: false
+  dtype: bfloat16
+  llm:
+    dimin: 2304
+  num_seq_per_batch: 100
+  num_total_steps: 2000
+  use: true
+  wandb_project_name: TemporalSAE_test
+sae:
+  block_id: 1
+  bottleneck_factor: 1
+  exp_factor: 4
+  gamma_reg: 10
+  kval_topk: 100
+  n_attn_layers: 1
+  n_heads: 4
+  normalize_weights: false
+  sae_diff_type: topk
+  tied_weights: true
+seed: 42
+tag: scratch

klbxe18a/latest_ckpt.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a7c2cc6c4a5b1b178bde8d7eb954a67c3a90af69ae601d880d85ed4b409085b
+size 4332159098