diff options
Diffstat (limited to 'runs-take2/rnn')
| -rw-r--r-- | runs-take2/rnn/seed0/best.pt | bin | 0 -> 150595 bytes | |||
| -rw-r--r-- | runs-take2/rnn/seed0/config.json | 30 | ||||
| -rw-r--r-- | runs-take2/rnn/seed0/metrics.csv | 23 | ||||
| -rw-r--r-- | runs-take2/rnn/seed0/run_meta.json | 14 |
4 files changed, 67 insertions, 0 deletions
diff --git a/runs-take2/rnn/seed0/best.pt b/runs-take2/rnn/seed0/best.pt Binary files differnew file mode 100644 index 0000000..3dc0b21 --- /dev/null +++ b/runs-take2/rnn/seed0/best.pt diff --git a/runs-take2/rnn/seed0/config.json b/runs-take2/rnn/seed0/config.json new file mode 100644 index 0000000..5433dd4 --- /dev/null +++ b/runs-take2/rnn/seed0/config.json @@ -0,0 +1,30 @@ +{ + "vocab_mode": "digits", + "range_start": 2, + "range_end": 100, + "holdout_frac": 0.3, + "seed": 0, + "model": "rnn", + "d_model": 128, + "max_steps": 20, + "halting": true, + "halt_penalty": 0.01, + "halt_warmup_steps": 1000, + "halt_ramp_end_steps": 5000, + "min_steps": 2, + "n_layers": 2, + "n_heads": 4, + "lr": 0.001, + "weight_decay": 1.0, + "max_train_steps": 200000, + "eval_every": 200, + "early_stop_em": 1.0, + "early_stop_patience": 5, + "batch_size": 32, + "max_out_len": 6, + "log_n_examples": 10, + "out_dir": "runs-take2", + "vocab": 11, + "eos_id": 10, + "pad_id": 11 +}
\ No newline at end of file diff --git a/runs-take2/rnn/seed0/metrics.csv b/runs-take2/rnn/seed0/metrics.csv new file mode 100644 index 0000000..9269796 --- /dev/null +++ b/runs-take2/rnn/seed0/metrics.csv @@ -0,0 +1,23 @@ +step,train_loss,train_token_acc,train_em,val_token_acc,val_em,mean_halt_steps,log_examples,param_count
+200,0.4903171956539154,0.8701923076923077,0.6666666666666666,0.5340909090909091,0.1,4.234026908874512,3:3~5;4:7~5;8:9~11;9:1~11;10:ok;12:33~13;13:11~17;15:33~17;18:31~19;21:33~23,36620
+400,0.17726276814937592,0.9855769230769231,0.9565217391304348,0.5681818181818182,0.2,6.274046421051025,3:3~5;4:7~5;8:9~11;9:1~11;10:ok;12:33~13;13:11~17;15:37~17;18:31~19;21:ok,36620
+600,0.5476387143135071,0.9278846153846154,0.782608695652174,0.5909090909090909,0.2,7.251992225646973,3:3~5;4:7~5;8:9~11;9:1~11;10:1~11;12:33~13;13:11~17;15:37~17;18:31~19;21:ok,36620
+800,0.09844685345888138,1.0,1.0,0.5568181818181818,0.16666666666666666,7.7551164627075195,3:3~5;4:7~5;8:9~11;9:1~11;10:ok;12:23~13;13:31~17;15:37~17;18:31~19;21:ok,36620
+1000,0.10235479474067688,1.0,1.0,0.5909090909090909,0.13333333333333333,10.119675636291504,3:7~5;4:7~5;8:9~11;9:1~11;10:ok;12:23~13;13:11~17;15:73~17;18:31~19;21:ok,36620
+1200,0.26242315769195557,0.9855769230769231,0.9565217391304348,0.6022727272727273,0.2,6.777245998382568,3:3~5;4:7~5;8:9~11;9:1~11;10:ok;12:23~13;13:13~17;15:37~17;18:31~19;21:ok,36620
+1400,0.29470914602279663,0.8990384615384616,0.7681159420289855,0.5340909090909091,0.16666666666666666,8.125110626220703,3:7~5;4:7~5;8:9~11;9:ok;10:1~11;12:23~13;13:11~17;15:23~17;18:23~19;21:ok,36620
+1600,0.14815756678581238,0.9903846153846154,0.9710144927536232,0.6022727272727273,0.2,8.417304992675781,3:3~5;4:7~5;8:1~11;9:1~11;10:ok;12:23~13;13:13~17;15:79~17;18:21~19;21:ok,36620
+1800,0.14240387082099915,0.9951923076923077,0.9855072463768116,0.6136363636363636,0.23333333333333334,10.154568672180176,3:7~5;4:7~5;8:9~11;9:1~11;10:ok;12:23~13;13:11~17;15:77~17;18:31~19;21:ok,36620
+2000,0.34200283885002136,0.9182692307692307,0.7681159420289855,0.6363636363636364,0.3,7.1683454513549805,3:7~5;4:7~5;8:9~11;9:1~11;10:ok;12:23~13;13:ok;15:77~17;18:23~19;21:ok,36620
+2200,0.1432158350944519,0.9855769230769231,0.9565217391304348,0.6931818181818182,0.3333333333333333,10.690964698791504,3:7~5;4:7~5;8:1~11;9:1~11;10:ok;12:23~13;13:11~17;15:97~17;18:29~19;21:ok,36620
+2400,0.35249558091163635,0.9326923076923077,0.855072463768116,0.5795454545454546,0.16666666666666666,8.192712783813477,3:3~5;4:7~5;8:1~11;9:1~11;10:ok;12:23~13;13:11~17;15:59~17;18:21~19;21:ok,36620
+2600,0.23677189648151398,0.9807692307692307,0.9420289855072463,0.6022727272727273,0.16666666666666666,9.858160018920898,3:3~5;4:7~5;8:1~11;9:1~11;10:ok;12:23~13;13:11~17;15:77~17;18:23~19;21:ok,36620
+2800,0.34069204330444336,0.9423076923076923,0.8260869565217391,0.5568181818181818,0.13333333333333333,6.819947242736816,3:3~5;4:7~5;8:9~11;9:1~11;10:ok;12:29~13;13:11~17;15:59~17;18:83~19;21:29~23,36620
+3000,0.24851882457733154,0.9230769230769231,0.8115942028985508,0.7045454545454546,0.36666666666666664,6.370993614196777,3:3~5;4:7~5;8:9~11;9:7~11;10:ok;12:23~13;13:37~17;15:71~17;18:33~19;21:ok,36620
+3200,0.38633477687835693,0.9519230769230769,0.8840579710144928,0.6136363636363636,0.26666666666666666,6.982702255249023,3:3~5;4:7~5;8:7~11;9:1~11;10:ok;12:23~13;13:31~17;15:79~17;18:83~19;21:ok,36620
+3400,0.22584262490272522,1.0,1.0,0.6477272727272727,0.23333333333333334,8.568931579589844,3:7~5;4:7~5;8:1~11;9:1~11;10:ok;12:23~13;13:37~17;15:79~17;18:23~19;21:ok,36620
+3600,0.6283265948295593,0.8894230769230769,0.7391304347826086,0.5681818181818182,0.2,8.007800102233887,3:3~5;4:7~5;8:7~11;9:3~11;10:ok;12:33~13;13:11~17;15:73~17;18:83~19;21:ok,36620
+3800,0.3623741865158081,0.9423076923076923,0.855072463768116,0.6818181818181818,0.3333333333333333,7.691666603088379,3:3~5;4:7~5;8:1~11;9:3~11;10:ok;12:23~13;13:31~17;15:97~17;18:29~19;21:ok,36620
+4000,0.2987810969352722,0.9807692307692307,0.9420289855072463,0.6022727272727273,0.2,7.724488735198975,3:3~5;4:7~5;8:3~11;9:1~11;10:ok;12:23~13;13:11~17;15:79~17;18:33~19;21:ok,36620
+4200,0.2570030093193054,0.9375,0.8405797101449275,0.5909090909090909,0.2,8.140658378601074,3:3~5;4:7~5;8:8~11;9:1~11;10:ok;12:23~13;13:37~17;15:97~17;18:83~19;21:29~23,36620
+4400,1.5922589302062988,0.6153846153846154,0.2463768115942029,0.4772727272727273,0.06666666666666667,5.81683349609375,3:7~5;4:7~5;8:-1~11;9:1~11;10:71~11;12:33~13;13:79~17;15:73~17;18:89~19;21:33~23,36620
diff --git a/runs-take2/rnn/seed0/run_meta.json b/runs-take2/rnn/seed0/run_meta.json new file mode 100644 index 0000000..587b392 --- /dev/null +++ b/runs-take2/rnn/seed0/run_meta.json @@ -0,0 +1,14 @@ +{ + "python": "3.11.15", + "torch": "2.13.0+cpu", + "numpy": "2.4.6", + "device": "cpu", + "torch_threads": 2, + "cmd": [ + "/home/jayrup/Projects/prime-grokking/src/train.py", + "rnn", + "0", + "--out_dir", + "runs-take2" + ] +}
\ No newline at end of file |
