Upload folder using huggingface_hub
Browse files- cfg.json +1 -0
- sae_weights.safetensors +3 -0
- sparsity.safetensors +3 -0
cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"model_name": "llava-hf/llava-v1.6-mistral-7b-hf", "local_model_path": "/data/changye/llava-hf/llava-v1.6-mistral-7b-hf", "model_class_name": "HookedLlava", "hook_name": "blocks.8.hook_resid_post", "hook_eval": "NOT_IN_USE", "hook_layer": 8, "hook_head_index": null, "dataset_path": "/data/changye/obelic10k-tokenized-text-llava", "dataset_trust_remote_code": true, "streaming": true, "is_dataset_tokenized": true, "context_size": 2048, "use_cached_activations": false, "cached_activations_path": null, "architecture": "standard", "d_in": 4096, "d_sae": 65536, "b_dec_init_method": "zeros", "expansion_factor": 16, "activation_fn": "relu", "activation_fn_kwargs": {}, "normalize_sae_decoder": false, "noise_scale": 0.0, "from_pretrained_path": null, "apply_b_dec_to_input": false, "decoder_orthogonal_init": false, "decoder_heuristic_init": true, "init_encoder_as_decoder_transpose": true, "n_batches_in_buffer": 32, "training_tokens": 163840000, "finetuning_tokens": 0, "store_batch_size_prompts": 4, "train_batch_size_tokens": 8192, "normalize_activations": "expected_average_only_in", "device": "cuda:7", "act_store_device": "cuda:7", "seed": 42, "dtype": "float32", "prepend_bos": true, "autocast": false, "autocast_lm": false, "compile_llm": false, "llm_compilation_mode": null, "compile_sae": false, "sae_compilation_mode": null, "adam_beta1": 0.9, "adam_beta2": 0.999, "mse_loss_normalization": null, "l1_coefficient": 5, "lp_norm": 1.0, "scale_sparsity_penalty_by_decoder_norm": true, "l1_warm_up_steps": 1000, "lr": 5e-05, "lr_scheduler_name": "constant", "lr_warm_up_steps": 0, "lr_end": 5e-06, "lr_decay_steps": 4000, "n_restart_cycles": 1, "finetuning_method": null, "use_ghost_grads": false, "feature_sampling_window": 1000, "dead_feature_window": 1000, "dead_feature_threshold": 0.0001, "n_eval_batches": 10, "eval_batch_size_prompts": null, "log_to_wandb": true, "log_activations_store_to_wandb": false, "log_optimizer_state_to_wandb": false, "wandb_project": "interp", "wandb_id": null, "run_name": "65536-L1-5-LR-5e-05-Tokens-1.638e+08", "wandb_entity": null, "wandb_log_frequency": 30, "eval_every_n_wandb_logs": 20, "resume": false, "n_checkpoints": 20, "checkpoint_path": "checkpoints/79i0ar03", "verbose": true, "model_kwargs": {}, "model_from_pretrained_kwargs": {"n_devices": 7}, "sae_lens_version": "3.20.0", "sae_lens_training_version": "3.20.0", "tokens_per_buffer": 536870912}
|
sae_weights.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0b834807d1e786e49d0144641dda57fbaa76e4e365dd2db850df2c3f1e44b087
|
3 |
+
size 2147762504
|
sparsity.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:70f33397d31e773e051c2b5d909fdda759765f113fc4a4cc9d5050aab280a9f3
|
3 |
+
size 262224
|