diff --git a/transcoder/T0_S-1_R1_P0.pt b/transcoder/T0_S-1_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..9e8f8c1f3cdd0f26cf69960f050ea83ebf8d784c --- /dev/null +++ b/transcoder/T0_S-1_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66858ff63daec87bea9857e71c92e8c6d9552ce9c6731757a1c0a1d3ea70998d +size 153705080 diff --git a/transcoder/T0_S-1_R1_P0_config.json b/transcoder/T0_S-1_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..b8c738c845de5aebd0dd286089845d353d10c0b6 --- /dev/null +++ b/transcoder/T0_S-1_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 0, + "l1_exp": -1, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M0_S-1_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T0_S-2_R1_P0.pt b/transcoder/T0_S-2_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..a9e29134bfe72580e150d2e23556b75f1bd942d1 --- /dev/null +++ b/transcoder/T0_S-2_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0dfedcf4be35d1b44c3dbac89466a2c9dbec2f1b7cdbbbc297c975d57073fd6 +size 153705080 diff --git a/transcoder/T0_S-2_R1_P0_config.json b/transcoder/T0_S-2_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..e939e3448e78bc204729a544e9fe3e30d63ef681 --- /dev/null +++ b/transcoder/T0_S-2_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 0, + "l1_exp": -2, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M0_S-2_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T0_S-3_R1_P0.pt b/transcoder/T0_S-3_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..2fea430b529efee597f7dca8288d60f8877806c3 --- /dev/null +++ b/transcoder/T0_S-3_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:195a6971c8889da0912b9997f783a789e5ce87984dac0d2a7cb2d7b35ac8047e +size 153705080 diff --git a/transcoder/T0_S-3_R1_P0_config.json b/transcoder/T0_S-3_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2c3a248215c1b6906d19af61acba3f7afa9d57cc --- /dev/null +++ b/transcoder/T0_S-3_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 0, + "l1_exp": -3, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M0_S-3_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T0_S-4_R1_P0.pt b/transcoder/T0_S-4_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..888fef16dc546f4424707e39b704c86acde75dcd --- /dev/null +++ b/transcoder/T0_S-4_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9efbe08eb25cd2666f7207822974318176fe4ccd2fbdd9e5319adfc17519f496 +size 153705080 diff --git a/transcoder/T0_S-4_R1_P0_config.json b/transcoder/T0_S-4_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3286996a0ed480d731993a94286f68da62c24ee8 --- /dev/null +++ b/transcoder/T0_S-4_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 0, + "l1_exp": -4, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M0_S-4_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T0_S-5_R1_P0.pt b/transcoder/T0_S-5_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..16d16d22169279b88ebf18c8a08c821f3eb6704d --- /dev/null +++ b/transcoder/T0_S-5_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8efa46999986a8ed7a561c74ca0acdc6019b73963a1795823d7869264f07964d +size 153705080 diff --git a/transcoder/T0_S-5_R1_P0_config.json b/transcoder/T0_S-5_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..505451d250bd62b5470c3068b2f84df726e6be38 --- /dev/null +++ b/transcoder/T0_S-5_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 0, + "l1_exp": -5, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M0_S-5_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T0_S-6_R1_P0.pt b/transcoder/T0_S-6_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..e7d64bb3694cd3262ececaf71f6326a11d1349db --- /dev/null +++ b/transcoder/T0_S-6_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1a763ed8e7d9728d6bf2ba5f48c677663eda1aee0ca1175d3d493d3db689b5f6 +size 153705080 diff --git a/transcoder/T0_S-6_R1_P0_config.json b/transcoder/T0_S-6_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..233d387f43c2c76c56bcaa91ba0f0130a652e3d1 --- /dev/null +++ b/transcoder/T0_S-6_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 0, + "l1_exp": -6, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M0_S-6_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T0_S-8_R1_P0.pt b/transcoder/T0_S-8_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..0f94c3118f8fc8bf9b57b71ddff0bf13518b9edc --- /dev/null +++ b/transcoder/T0_S-8_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ea3673ef5475aaa29ebfbe4ffc567a0e4e02210e57f54c8c533c8a58f2aba34 +size 153705080 diff --git a/transcoder/T0_S-8_R1_P0_config.json b/transcoder/T0_S-8_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..b7c1f04e9daff965274293132007efbe87d14921 --- /dev/null +++ b/transcoder/T0_S-8_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 0, + "l1_exp": -8, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M0_S-8_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T0_S0_R1_P0.pt b/transcoder/T0_S0_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..5ba8f6b5b8a288329fd5c053b18ffc63c90032cd --- /dev/null +++ b/transcoder/T0_S0_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96de880e9cdb53b3a3e906cb7668b4919c3e9f7aca8b70e73f4f2c47b1a1bb60 +size 153705072 diff --git a/transcoder/T0_S0_R1_P0_config.json b/transcoder/T0_S0_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..e02ccbe369a5d995f34e856864d983123bd776f4 --- /dev/null +++ b/transcoder/T0_S0_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 0, + "l1_exp": 0, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M0_S0_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T1_S-1_R1_P0.pt b/transcoder/T1_S-1_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..be359b8fcbf2de02576fac36c0127cea13804bbb --- /dev/null +++ b/transcoder/T1_S-1_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:923f248312ed5dbda72f0fa8f04a849a6e53aa3f5d0fa1d1ceda644f1a7c4717 +size 153705080 diff --git a/transcoder/T1_S-1_R1_P0_config.json b/transcoder/T1_S-1_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..536dc43eb1073eb908c17d633dcecb7296352137 --- /dev/null +++ b/transcoder/T1_S-1_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 1, + "l1_exp": -1, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M1_S-1_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T1_S-2_R1_P0.pt b/transcoder/T1_S-2_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..ce71a7eccdb625dc674381887724e7b2d447329e --- /dev/null +++ b/transcoder/T1_S-2_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0497a41e4b896bd13be23befb53789323358189c10a3e342147f9a82e7a94ca9 +size 153705080 diff --git a/transcoder/T1_S-2_R1_P0_config.json b/transcoder/T1_S-2_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d1c582e657ea3813fc0e9cdd344fde0737dfa8d0 --- /dev/null +++ b/transcoder/T1_S-2_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 1, + "l1_exp": -2, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M1_S-2_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T1_S-3_R1_P0.pt b/transcoder/T1_S-3_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..2cbc0b2f50fc6510e5757001d16832fd90b0154a --- /dev/null +++ b/transcoder/T1_S-3_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52798c267acaf04f8ee4346cdd7865e178614cb43cd3d062d549949f47bd0f3e +size 153705080 diff --git a/transcoder/T1_S-3_R1_P0_config.json b/transcoder/T1_S-3_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2e03de9d1e58e4779c452703a57550c78fe215d4 --- /dev/null +++ b/transcoder/T1_S-3_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 1, + "l1_exp": -3, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M1_S-3_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T1_S-4_R1_P0.pt b/transcoder/T1_S-4_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..1da0e579f93741133e5c5abb98216b11516d3efc --- /dev/null +++ b/transcoder/T1_S-4_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ca8d3112dc82cdca7fdc2ce5d058afc22e1cec117c0de3ec061e100b2afa8da +size 153705080 diff --git a/transcoder/T1_S-4_R1_P0_config.json b/transcoder/T1_S-4_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..76e2ea594b9638ec60195c1ef4ddc6ad6e2ccd37 --- /dev/null +++ b/transcoder/T1_S-4_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 1, + "l1_exp": -4, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M1_S-4_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T1_S-5_R1_P0.pt b/transcoder/T1_S-5_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc18c9860186227fd6efe87d257d41d75016b137 --- /dev/null +++ b/transcoder/T1_S-5_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:93500ea0409d5b3d6dd5bc03e3cc812c136a87516ea3e9a16058454ba0c0e88f +size 153705080 diff --git a/transcoder/T1_S-5_R1_P0_config.json b/transcoder/T1_S-5_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a72ef0148423b3626f6b7c2acb634ad8f81400aa --- /dev/null +++ b/transcoder/T1_S-5_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 1, + "l1_exp": -5, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M1_S-5_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T1_S-6_R1_P0.pt b/transcoder/T1_S-6_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..90d6cec1dd823148685afe47a73133f93d468ece --- /dev/null +++ b/transcoder/T1_S-6_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:729c1ee01a36e69f7aaad847837e7134b157188f882529ffdbbd2486c6438b2f +size 153705080 diff --git a/transcoder/T1_S-6_R1_P0_config.json b/transcoder/T1_S-6_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..acbdd69b5fbc1a9a7257192c7b243ba4ec78dc43 --- /dev/null +++ b/transcoder/T1_S-6_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 1, + "l1_exp": -6, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M1_S-6_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T1_S0_R1_P0.pt b/transcoder/T1_S0_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..3d1864223a1d34229864936c6792e99e56ec83aa --- /dev/null +++ b/transcoder/T1_S0_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e1f097e45bd999c7199911199e40be6f954d29bd25a94d4437bb555ab04b4561 +size 153705072 diff --git a/transcoder/T1_S0_R1_P0_config.json b/transcoder/T1_S0_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3d18945dec2c83695db7f35b4eb7c41e64df8871 --- /dev/null +++ b/transcoder/T1_S0_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 1, + "l1_exp": 0, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M1_S0_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T2_S-1_R1_P0.pt b/transcoder/T2_S-1_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..765fe055d0542ced34ca328900fd3bf670a1c05c --- /dev/null +++ b/transcoder/T2_S-1_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:770c41ccb13f03301332d27b886975e4d92d5e2bb1599d0aa8325f0a16c17905 +size 153705080 diff --git a/transcoder/T2_S-1_R1_P0_config.json b/transcoder/T2_S-1_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..6e2c995e2a18e2e50b9946aa7b357f79d2812559 --- /dev/null +++ b/transcoder/T2_S-1_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 2, + "l1_exp": -1, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M2_S-1_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T2_S-2_R1_P0.pt b/transcoder/T2_S-2_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..434f9d8ca2f83fca964a2a5744d6b50d3e643192 --- /dev/null +++ b/transcoder/T2_S-2_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c739e3b372bc2d7eeec1d136c43324e79836bd87b5ae72b83adba6faec0a363 +size 153705080 diff --git a/transcoder/T2_S-2_R1_P0_config.json b/transcoder/T2_S-2_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2a9d740f71f43451f1e7b6399962af7a07771156 --- /dev/null +++ b/transcoder/T2_S-2_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 2, + "l1_exp": -2, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M2_S-2_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T2_S-3_R1_P0.pt b/transcoder/T2_S-3_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..71a6dbb0311efd4e9f4e30f929ea743cabb185aa --- /dev/null +++ b/transcoder/T2_S-3_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea6f86cb420033cdfbdfd8ec33518305fa7b415b89fa7c0f9b5c6a04891caf44 +size 153705080 diff --git a/transcoder/T2_S-3_R1_P0_config.json b/transcoder/T2_S-3_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..af50942fc18ad7da0d8eb9a9d5b5dae94aad3a8b --- /dev/null +++ b/transcoder/T2_S-3_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 2, + "l1_exp": -3, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M2_S-3_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T2_S-4_R1_P0.pt b/transcoder/T2_S-4_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..639fcbfda4d1ba4a95ef91e9625f6294d66a2109 --- /dev/null +++ b/transcoder/T2_S-4_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d718ab852d85f1e221ed43d7f0859307cb8c3ef1ac9d334f677b8d7f7bf7f233 +size 153705080 diff --git a/transcoder/T2_S-4_R1_P0_config.json b/transcoder/T2_S-4_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..802e7dcc5fba154e7d227a61f78fd69262cbf3ff --- /dev/null +++ b/transcoder/T2_S-4_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 2, + "l1_exp": -4, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M2_S-4_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T2_S-5_R1_P0.pt b/transcoder/T2_S-5_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..05f807d6c9eaedeab137322eba3d59632c8c3ef6 --- /dev/null +++ b/transcoder/T2_S-5_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cbea78a42e417b7bbc6571641dab05735c6be4ed588c17e369f6462040d11d3b +size 153705080 diff --git a/transcoder/T2_S-5_R1_P0_config.json b/transcoder/T2_S-5_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..511ecfd323d09d92683988709b4ff99438b9a319 --- /dev/null +++ b/transcoder/T2_S-5_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 2, + "l1_exp": -5, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M2_S-5_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T2_S-6_R1_P0.pt b/transcoder/T2_S-6_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..9aa527cbefbb151f7c6e800da84e95ec7c5dff42 --- /dev/null +++ b/transcoder/T2_S-6_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77229011a320dd8f990ce1bcd81b2a3227287f78f250c2c89b2e2dafbe9ca249 +size 153705080 diff --git a/transcoder/T2_S-6_R1_P0_config.json b/transcoder/T2_S-6_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..e3b605b48e8310f52c3afe07500d10a373580585 --- /dev/null +++ b/transcoder/T2_S-6_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 2, + "l1_exp": -6, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M2_S-6_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T2_S-8_R1_P0.pt b/transcoder/T2_S-8_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..32b773eacaccb912f32246f5f422a4ab9170769b --- /dev/null +++ b/transcoder/T2_S-8_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:76d8158a249dedcc50b2e714edf48d2372cb91a4c79be6fa726c5f2d793e5d5a +size 153705080 diff --git a/transcoder/T2_S-8_R1_P0_config.json b/transcoder/T2_S-8_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..96710f91c8cfe9d0773a2e3be10acb1de9bf460c --- /dev/null +++ b/transcoder/T2_S-8_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 2, + "l1_exp": -8, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M2_S-8_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T2_S0_R1_P0.pt b/transcoder/T2_S0_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..5e7b7a412e52f5c716a9dca137963558e90c22cb --- /dev/null +++ b/transcoder/T2_S0_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39bd095230c89baac55502b0aa0c95d1d89432a20c58c847387bf08536b363c7 +size 153705072 diff --git a/transcoder/T2_S0_R1_P0_config.json b/transcoder/T2_S0_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..8c21b52e78bd13602055e783400182e2b0d5550b --- /dev/null +++ b/transcoder/T2_S0_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 2, + "l1_exp": 0, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M2_S0_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T3_S-1_R1_P0.pt b/transcoder/T3_S-1_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ddef4250988cccd6cd73a8cb46a2abaa758087d --- /dev/null +++ b/transcoder/T3_S-1_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e40312c27b9e817fa4881f542d0e2545df0768df4fab1e27803713d61e78d5df +size 153705080 diff --git a/transcoder/T3_S-1_R1_P0_config.json b/transcoder/T3_S-1_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..b81e50748cb39361a3b4f89b3f0f779e42a12440 --- /dev/null +++ b/transcoder/T3_S-1_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 3, + "l1_exp": -1, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M3_S-1_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T3_S-2_R1_P0.pt b/transcoder/T3_S-2_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..dab953516ed95c01ad4de884650cce022591720f --- /dev/null +++ b/transcoder/T3_S-2_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15b24e29464fd7da37c9ecccbd611e07a5233f01402a7f643bc5012885cce70a +size 153705080 diff --git a/transcoder/T3_S-2_R1_P0_config.json b/transcoder/T3_S-2_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2a124cab434e45246982df19dcaf911eacc41eba --- /dev/null +++ b/transcoder/T3_S-2_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 3, + "l1_exp": -2, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M3_S-2_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T3_S-3_R1_P0.pt b/transcoder/T3_S-3_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..a0459128e3dbfd4130587e3838c90e807d3d87d5 --- /dev/null +++ b/transcoder/T3_S-3_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3a6310193d81562b33cb159c5a7bb28033ed66314c38234aada580ab46dd560 +size 153705080 diff --git a/transcoder/T3_S-3_R1_P0_config.json b/transcoder/T3_S-3_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c328a724f4864999e50c0f76acf811e54b8fe2bb --- /dev/null +++ b/transcoder/T3_S-3_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 3, + "l1_exp": -3, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M3_S-3_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T3_S-4_R1_P0.pt b/transcoder/T3_S-4_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..604d478ab9d68e05eb0eb9cca4c91b6f2cb0b746 --- /dev/null +++ b/transcoder/T3_S-4_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:deae00716bf25da3d4a852ed3fbde32a919e769517239995a00d7578cf0a291a +size 153705080 diff --git a/transcoder/T3_S-4_R1_P0_config.json b/transcoder/T3_S-4_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..b4e7047234bcc01118bde4f903ba184bf490d64b --- /dev/null +++ b/transcoder/T3_S-4_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 3, + "l1_exp": -4, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M3_S-4_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T3_S-5_R1_P0.pt b/transcoder/T3_S-5_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..5de2968287a519b965db7fc7e58a0bc62d22193a --- /dev/null +++ b/transcoder/T3_S-5_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b21ba9fc973fc93650743be9af060a63dd2b0b9fa8dec9304a8e24bd508ce256 +size 153705080 diff --git a/transcoder/T3_S-5_R1_P0_config.json b/transcoder/T3_S-5_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd8ae42e36e3c8c8c074439cbd9f9b94547c9bb8 --- /dev/null +++ b/transcoder/T3_S-5_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 3, + "l1_exp": -5, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M3_S-5_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T3_S-6_R1_P0.pt b/transcoder/T3_S-6_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..18870d51a49d72f3bc90defd9a31c4f2a21ed707 --- /dev/null +++ b/transcoder/T3_S-6_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb52a39b2c188f6b0f7e75a7d99970995305129e1c18b6a7fba8fc9114eec192 +size 153705080 diff --git a/transcoder/T3_S-6_R1_P0_config.json b/transcoder/T3_S-6_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..897b7a131fb67145d7bc6cd5b584d28ccce6d0d0 --- /dev/null +++ b/transcoder/T3_S-6_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 3, + "l1_exp": -6, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M3_S-6_R1_P0" +} \ No newline at end of file diff --git a/transcoder/T3_S0_R1_P0.pt b/transcoder/T3_S0_R1_P0.pt new file mode 100644 index 0000000000000000000000000000000000000000..d9487f444a8696156a7144aa6d833496f4c30205 --- /dev/null +++ b/transcoder/T3_S0_R1_P0.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d5055b197e78cfc1cbf7dd43ab21cc454fa4d998a8866050c6ea4fb57af62ca +size 153705072 diff --git a/transcoder/T3_S0_R1_P0_config.json b/transcoder/T3_S0_R1_P0_config.json new file mode 100644 index 0000000000000000000000000000000000000000..945905bf9d6bcc1b1e77d80350d8e9e831fac014 --- /dev/null +++ b/transcoder/T3_S0_R1_P0_config.json @@ -0,0 +1,38 @@ +{ + "n_features": 25000, + "d_model": 768, + "lr_exp": -10, + "disable_comet": false, + "per_neuron_reinit_interval": 0, + "reservoir_time_discount": 0.995, + "reinit_interval": 800, + "max_reinit_neurons": 5000, + "reservoir_size": 5000, + "n_piles": 292, + "log_interval": 200, + "reinit_input_norm": "target_scaled", + "reinit_input": "x", + "reinit_norm_alpha": 0.3, + "data_loc": "mlp_data", + "reinit_threshold": -6, + "scheduler": "wsd", + "layer_idx": 3, + "l1_exp": 0, + "neuron_reinit_percent": 0.85, + "beta1": 1, + "beta2": 4, + "reinit_target": "y", + "sparse_adam": false, + "run_template": "M{layer_idx}_S{l1_exp}_R{l1_ratio}_P{l1_p}", + "project_name": "mlp_map_test", + "decoder_bias": true, + "l1_beta": 0.99, + "alt_sparsity_loss": "log", + "l1_ratio": 1, + "l1_p": 0, + "optimizer": "sparse_adam", + "model_type": "mlp_map", + "adam_beta1": 0.5, + "adam_beta2": 0.9375, + "run_name": "M3_S0_R1_P0" +} \ No newline at end of file