diff --git a/.gitattributes b/.gitattributes
index a6344aac8c09253b3b630fb776ae94478aa0275b..38a601b8e09e0ecccdb11292fadcb9d2e98abb1e 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -33,3 +33,28 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+empirical_global_heatmaps/empirical_global_raw.png filter=lfs diff=lfs merge=lfs -text
+empirical_global_heatmaps/empirical_global_sim.png filter=lfs diff=lfs merge=lfs -text
+empirical_l1_head_heatmaps/local_head0_vs_head0.png filter=lfs diff=lfs merge=lfs -text
+empirical_l1_head_heatmaps/local_head0_vs_head1.png filter=lfs diff=lfs merge=lfs -text
+empirical_l1_head_heatmaps/local_head1_vs_head0.png filter=lfs diff=lfs merge=lfs -text
+empirical_l1_head_heatmaps/local_head1_vs_head1.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_direct_vs_direct.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_direct_vs_layer1.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_direct_vs_layer2.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_layer1_vs_direct.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_layer1_vs_layer1.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_layer1_vs_layer2.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_layer2_vs_direct.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_layer2_vs_layer1.png filter=lfs diff=lfs merge=lfs -text
+empirical_layer_block_heatmaps/local_layer2_vs_layer2.png filter=lfs diff=lfs merge=lfs -text
+global_tn_heatmaps/global_tn_sim.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_direct.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_layer1.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_layer2.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_direct.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_layer1.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_layer2.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_direct.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_layer1.png filter=lfs diff=lfs merge=lfs -text
+path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_layer2.png filter=lfs diff=lfs merge=lfs -text
diff --git a/checkpoints/step_0.pt b/checkpoints/step_0.pt
new file mode 100644
index 0000000000000000000000000000000000000000..7f2f33a166eda4d800cb881ffd34c4e66869c43e
--- /dev/null
+++ b/checkpoints/step_0.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7a9bee96061702fe4b9d5e6de6b59a286a8a05e554f452c56b310d441097395
+size 14261
diff --git a/checkpoints/step_1020.pt b/checkpoints/step_1020.pt
new file mode 100644
index 0000000000000000000000000000000000000000..88fda7bc9a578374d4f9b1b5b10af3f6ee7a9a6b
--- /dev/null
+++ b/checkpoints/step_1020.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46d0606595eb46020199e1ab49dc1c0ebf4bff70e46ba9fcc243f373200185a5
+size 14949
diff --git a/checkpoints/step_1080.pt b/checkpoints/step_1080.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6cedd082dc4c96540cd17aa9b4cbcc2318d7bb78
--- /dev/null
+++ b/checkpoints/step_1080.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a3222d118db5694e2d2396f95cb20be5baba719e04a47015d351665ac5aafb3
+size 14949
diff --git a/checkpoints/step_1140.pt b/checkpoints/step_1140.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a48f82b16e3a4c5e7eec24c7bdcaa00f62a9655f
--- /dev/null
+++ b/checkpoints/step_1140.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:738888683be7ca4a5e8444c47b468742f9007ee4c1d5bb728bc3694cb672e9dd
+size 14949
diff --git a/checkpoints/step_120.pt b/checkpoints/step_120.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c78dc8e1853b6a73e9d921a43fd8f4258d894c42
--- /dev/null
+++ b/checkpoints/step_120.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ffed7e5f9df7cea982d6922d77aa6877282c830e0351c44f30a43955af68c9f
+size 14933
diff --git a/checkpoints/step_1200.pt b/checkpoints/step_1200.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0e151953fc87b2ae3699d7ca8bc7783c9b500740
--- /dev/null
+++ b/checkpoints/step_1200.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a99e4ea93cbf20225ad0b4bb4923bf6a8464d9af78d269da1e949c16031c7c8
+size 14949
diff --git a/checkpoints/step_1260.pt b/checkpoints/step_1260.pt
new file mode 100644
index 0000000000000000000000000000000000000000..86fd9e3ce213784ca9e1862164b0ed40278d6a19
--- /dev/null
+++ b/checkpoints/step_1260.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3046bfe58679eb1ce8cfa952b9f1b9a563d3a7f4cf624a1773a557884963ad97
+size 14949
diff --git a/checkpoints/step_1320.pt b/checkpoints/step_1320.pt
new file mode 100644
index 0000000000000000000000000000000000000000..9c7319356f3c2df0c969435322d79836a3da1b3d
--- /dev/null
+++ b/checkpoints/step_1320.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62733374b94cd6299a12fcc52c8d0bd14e81d0d1f194cdfd7490daa3e024f00f
+size 14949
diff --git a/checkpoints/step_1380.pt b/checkpoints/step_1380.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8fec54e5f979a70f43ccdee745740acd36def25a
--- /dev/null
+++ b/checkpoints/step_1380.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd7de757426b76bb2763e9e4d00e45dccebaf439efd78027010a9a1bb74d5c9c
+size 14949
diff --git a/checkpoints/step_1440.pt b/checkpoints/step_1440.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a1a0916c46a2694da93a90d2f79c6f13470124d6
--- /dev/null
+++ b/checkpoints/step_1440.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19e919be28ef38e21a01292e6b12714fd3cb6607c6bb4b46aadb1b557dea777a
+size 14949
diff --git a/checkpoints/step_1500.pt b/checkpoints/step_1500.pt
new file mode 100644
index 0000000000000000000000000000000000000000..454780a3c700a61f121fda9ecb8c4719eaa0dbd7
--- /dev/null
+++ b/checkpoints/step_1500.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21fa6d4735761eb6491ebd833ee8b35cdb35586e870231613f86cef99be47d29
+size 14949
diff --git a/checkpoints/step_1560.pt b/checkpoints/step_1560.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1e2a3b3b5032368e0ea49c4f09cde11431026cb5
--- /dev/null
+++ b/checkpoints/step_1560.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fa04703ca4a12659ad5ad302c8d398b16f57a5d4d12d67a16bf72e6a01585fe
+size 14949
diff --git a/checkpoints/step_1620.pt b/checkpoints/step_1620.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b7ecfe5841e3a99ae46b563b797e4f48f1f701b6
--- /dev/null
+++ b/checkpoints/step_1620.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9a09f1fab8295eb9ac55132f68d9bb7409ddbee531d4bf22af9681baa2faa83
+size 14949
diff --git a/checkpoints/step_1680.pt b/checkpoints/step_1680.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c447cf68db2808ac813d99e8b513aefecf7a3246
--- /dev/null
+++ b/checkpoints/step_1680.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91425d41da432f39ff1d0346b7eb828496de7c9d4026652f8b7abf1f3ca57a4c
+size 14949
diff --git a/checkpoints/step_1740.pt b/checkpoints/step_1740.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ca5768a4e76157479bc9e25fda856f6e9cc755d7
--- /dev/null
+++ b/checkpoints/step_1740.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e219bb9e753d49f78a0f3bbe330e22d476c7c181c841daf7806c8da384baa8fb
+size 14949
diff --git a/checkpoints/step_180.pt b/checkpoints/step_180.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4cdfbc989556b1fb5dd6660bb11cc865be85170e
--- /dev/null
+++ b/checkpoints/step_180.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d618fe0bbf60248ba9951db26d07c9bb6d404c4f8d041e04de0a79c6463886f0
+size 14933
diff --git a/checkpoints/step_1800.pt b/checkpoints/step_1800.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0212a9a8c5e039eef67d2768b54aa387e281e53f
--- /dev/null
+++ b/checkpoints/step_1800.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:113ab38fe2121ed1b139db75f1a7eba084c7fad8ab6fdb87da75194e289c8a2d
+size 14949
diff --git a/checkpoints/step_1860.pt b/checkpoints/step_1860.pt
new file mode 100644
index 0000000000000000000000000000000000000000..290434ea433f2340d608b24323167aa31ba1a85b
--- /dev/null
+++ b/checkpoints/step_1860.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20364728d02a6c7acab7eae501c40ca5743f0c6894fe22636e76249a374fc5ef
+size 14949
diff --git a/checkpoints/step_1920.pt b/checkpoints/step_1920.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d33f6e01e87b3a5088b9555cdadf803862daa52e
--- /dev/null
+++ b/checkpoints/step_1920.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d025125168ffd8ccbf89e805f6ff7635a17d7766f28888eac8da4b155757441a
+size 14949
diff --git a/checkpoints/step_1980.pt b/checkpoints/step_1980.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2f6107cb592fa2aaca7fed9c692c4609c9c1045d
--- /dev/null
+++ b/checkpoints/step_1980.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e57c31ef1902205f031ad17f1b3cd71d74c46ca417d79745371b3b7bf4ecfa82
+size 14949
diff --git a/checkpoints/step_2040.pt b/checkpoints/step_2040.pt
new file mode 100644
index 0000000000000000000000000000000000000000..9f424e84e5e11aac3f7b95b6299e329d8010cc15
--- /dev/null
+++ b/checkpoints/step_2040.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25c755a08b617cede7cb2a7d0872e2f7678e4b7ed73205b187b51ccb2c29dd32
+size 14949
diff --git a/checkpoints/step_2100.pt b/checkpoints/step_2100.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e7772f7b437488117db2393cf045a19a315cf549
--- /dev/null
+++ b/checkpoints/step_2100.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fca6b641801e50bb34db80a4fc2581ceb7a3a004ebb1527d317299f929e0b252
+size 14949
diff --git a/checkpoints/step_2160.pt b/checkpoints/step_2160.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a29b4469a66228c94dbe2c6e465d279aa7e07534
--- /dev/null
+++ b/checkpoints/step_2160.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9e2c7f539d6a8984390fa7952d3ca2851bd409947be6b107811330dd6370d48
+size 14949
diff --git a/checkpoints/step_2220.pt b/checkpoints/step_2220.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8bee36d0bca005bfd2bb6672fd787273992c3c74
--- /dev/null
+++ b/checkpoints/step_2220.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35d40f211abe56df07ad69e5e22d19af718bddcd5abc62e08761907f880e2bc4
+size 14949
diff --git a/checkpoints/step_2280.pt b/checkpoints/step_2280.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c84d79ca264ae34c7a38c380f7d8392271361366
--- /dev/null
+++ b/checkpoints/step_2280.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:397ae4e85fa9b8782eb996f192944e4a8af419d190e24cea709bd7984a990ab6
+size 14949
diff --git a/checkpoints/step_2340.pt b/checkpoints/step_2340.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4f6c970a1e2a3bff67b23349d4c197d0939ddd04
--- /dev/null
+++ b/checkpoints/step_2340.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cc99dbfa9416d3dfa51712576385a20c0e6fdb3b9544f19a8c5df2ccb4b9103
+size 14949
diff --git a/checkpoints/step_240.pt b/checkpoints/step_240.pt
new file mode 100644
index 0000000000000000000000000000000000000000..74d14e95b26faa80a442a3cdbef54d008921ab15
--- /dev/null
+++ b/checkpoints/step_240.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd2ba5ded2a24d75b6aa1af668060e25d2c6201a6e6997fb41f4d410ebb4095d
+size 14933
diff --git a/checkpoints/step_2400.pt b/checkpoints/step_2400.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4c28a5b5e6ade7e48c9201406cbca09707532fa3
--- /dev/null
+++ b/checkpoints/step_2400.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b067f9bd21dfa29a0123ae8ccfbf2abfa6cee930ecda217224460f209bda7e8
+size 14949
diff --git a/checkpoints/step_2460.pt b/checkpoints/step_2460.pt
new file mode 100644
index 0000000000000000000000000000000000000000..38bbd88b03e75f7acc3b551a5246c126fb1d287f
--- /dev/null
+++ b/checkpoints/step_2460.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0319310cea95891cf85090a87411828a343174599567934c257bb228df1a7d3e
+size 14949
diff --git a/checkpoints/step_2520.pt b/checkpoints/step_2520.pt
new file mode 100644
index 0000000000000000000000000000000000000000..cad010d4e1ed0faabde927f94f003df34c5076ff
--- /dev/null
+++ b/checkpoints/step_2520.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb2486e211108e9cec2fd6758f11922944721bbca45d5564c5bc97aa82a3b539
+size 14949
diff --git a/checkpoints/step_2580.pt b/checkpoints/step_2580.pt
new file mode 100644
index 0000000000000000000000000000000000000000..61b02c7f06d5171fe82e756304256a4c1f2428f8
--- /dev/null
+++ b/checkpoints/step_2580.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fe14bf75fb393dfa385ae9fc3834aa961bfb226d1bf3053d8ef5cd9709d2bec
+size 14949
diff --git a/checkpoints/step_2640.pt b/checkpoints/step_2640.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b164c9fbb1dd50fd7548bf598b34b3dd7d85fab1
--- /dev/null
+++ b/checkpoints/step_2640.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80149bf30e5a6f1ed0c91fc7fb0499f410c12253fdd42db7f53d543fcfca9b7d
+size 14949
diff --git a/checkpoints/step_2700.pt b/checkpoints/step_2700.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3f2a28da5b131f51f1be80d5ee2ebf362ceebe76
--- /dev/null
+++ b/checkpoints/step_2700.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eab936a808a26d93ed7e6cfc2145a38b871d247db862c2a3526eea32f588d8b4
+size 14949
diff --git a/checkpoints/step_2760.pt b/checkpoints/step_2760.pt
new file mode 100644
index 0000000000000000000000000000000000000000..67975609d1804bea4aa1229d801f82ad1efe6d0d
--- /dev/null
+++ b/checkpoints/step_2760.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc98cf9e90e2bb45cc48b25623d22b8b481e6397aff0708e9a375bfe69fe5593
+size 14949
diff --git a/checkpoints/step_2820.pt b/checkpoints/step_2820.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1ade341fd792cd62d8b96030f425ea787a2cd4bd
--- /dev/null
+++ b/checkpoints/step_2820.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61a8200ea7f898956db2c3c88772987a1601101d8e1381b97ffbc8d71fafde15
+size 14949
diff --git a/checkpoints/step_2880.pt b/checkpoints/step_2880.pt
new file mode 100644
index 0000000000000000000000000000000000000000..92a4d6e51ec45ebca60f015930f267fa28ea4fb0
--- /dev/null
+++ b/checkpoints/step_2880.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:664295040180de62caa0a6b8819caa9d7918ffcd7a1e720dff24cdd4a8fb20cc
+size 14949
diff --git a/checkpoints/step_2940.pt b/checkpoints/step_2940.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1b280c07b746fe0b7836c7cc7e74ff9be926bc46
--- /dev/null
+++ b/checkpoints/step_2940.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6294176bfb8c5116f98e34fac315a67e2af1d0aeefe645b6ff64c3f20722caec
+size 14949
diff --git a/checkpoints/step_300.pt b/checkpoints/step_300.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3fbcb78c802035ad87433f0b1d4821ed6f2b2d57
--- /dev/null
+++ b/checkpoints/step_300.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a10a1344083775b79b06e87a89452ff871593606b7cc507f6747735a2fa65a45
+size 14933
diff --git a/checkpoints/step_3000.pt b/checkpoints/step_3000.pt
new file mode 100644
index 0000000000000000000000000000000000000000..33a3e51c15c8fc989a71c293c132d2b40a74cb23
--- /dev/null
+++ b/checkpoints/step_3000.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41e1825c3ea63c1feda60b77b12f7cd40288d2ac75ba4f323fdad067c7936d6b
+size 14949
diff --git a/checkpoints/step_3060.pt b/checkpoints/step_3060.pt
new file mode 100644
index 0000000000000000000000000000000000000000..14fe25829bb8095c0a2df1ccdb1267671f6fa054
--- /dev/null
+++ b/checkpoints/step_3060.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e3d0a6821c85ec93fca4672d5588f6c79dddbb6dd8ba89cd9b75104ab815a1d
+size 14949
diff --git a/checkpoints/step_3120.pt b/checkpoints/step_3120.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2061a0e315c73919e8ce7203f643c462ce51d294
--- /dev/null
+++ b/checkpoints/step_3120.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:314395d27ef236145b692231a81b86c0cd046b4a30d912ee96baf12d190a0477
+size 14949
diff --git a/checkpoints/step_3180.pt b/checkpoints/step_3180.pt
new file mode 100644
index 0000000000000000000000000000000000000000..cee5c40012ca35d40119fba2c65adde6886f6702
--- /dev/null
+++ b/checkpoints/step_3180.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:081bd82b30acc8fcd4a92e7bcbd91fa37b27bd9e210fb751285b23340703b855
+size 14949
diff --git a/checkpoints/step_3240.pt b/checkpoints/step_3240.pt
new file mode 100644
index 0000000000000000000000000000000000000000..efc56deeee508601728ba5a5e2fc15371e034a65
--- /dev/null
+++ b/checkpoints/step_3240.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7c55e484621ab6d4552a673fb852adece6700085f1d6f82058a47767eaa7beb
+size 14949
diff --git a/checkpoints/step_3300.pt b/checkpoints/step_3300.pt
new file mode 100644
index 0000000000000000000000000000000000000000..71ff28458b028ca0e0ddc3030219eb596877477d
--- /dev/null
+++ b/checkpoints/step_3300.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12e7d36683f14afd95df0f1d1c23536c08e014febe97a9b026241c510b24e912
+size 14949
diff --git a/checkpoints/step_3360.pt b/checkpoints/step_3360.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ef3a233aeddea70f541f2b70056915cb07ec23c9
--- /dev/null
+++ b/checkpoints/step_3360.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6d6a045a0d82dbc0ff485d88919a7d91951a3e7bc8ad7113a47a8ef7a752513
+size 14949
diff --git a/checkpoints/step_3420.pt b/checkpoints/step_3420.pt
new file mode 100644
index 0000000000000000000000000000000000000000..14bf386427461358cfc1961a9f71e05da7370024
--- /dev/null
+++ b/checkpoints/step_3420.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:589cd9b34a8b7b48badc947054ae03f004886075d434cb4ea9bb404d1f2fec9b
+size 14949
diff --git a/checkpoints/step_3480.pt b/checkpoints/step_3480.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c0598a3d0e0a287ac45869b319a3a180911388d6
--- /dev/null
+++ b/checkpoints/step_3480.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fc2d352da79ea9e8cccbe413d3a59a9f2b0e5d916c96778bb6609dc532d6d4b
+size 14949
diff --git a/checkpoints/step_3540.pt b/checkpoints/step_3540.pt
new file mode 100644
index 0000000000000000000000000000000000000000..43c7349cd0d1b016ae6c2774ded4d065b0547df9
--- /dev/null
+++ b/checkpoints/step_3540.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06cd29cee04aff3cc5d2f36411914a631ec8dfe8e253fbec49e70fd644d6a18e
+size 14949
diff --git a/checkpoints/step_360.pt b/checkpoints/step_360.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e30fa75b37fa35c218047b4f34dc246473203df6
--- /dev/null
+++ b/checkpoints/step_360.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e8a60b2e7f31b973bfcc1e6583a96ac6d9920fe25f73a4573ce5185122615a8
+size 14933
diff --git a/checkpoints/step_3600.pt b/checkpoints/step_3600.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c273c57c80345e966e2a2d2aa38a3fad36a52905
--- /dev/null
+++ b/checkpoints/step_3600.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83b276309dd6ea33eff6ff40b33c44c6dc635331e3246472873c6a427da8448e
+size 14949
diff --git a/checkpoints/step_3660.pt b/checkpoints/step_3660.pt
new file mode 100644
index 0000000000000000000000000000000000000000..389ab003422e5cc75e8227f8de23c5c8bbf48dd6
--- /dev/null
+++ b/checkpoints/step_3660.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fce165c51b3d24298c7aba5ca767f90d128170986d68f8c10e0c3f2dc91f2289
+size 14949
diff --git a/checkpoints/step_3720.pt b/checkpoints/step_3720.pt
new file mode 100644
index 0000000000000000000000000000000000000000..85984fd71ea8f108bd665a6e7c60fdef40f05076
--- /dev/null
+++ b/checkpoints/step_3720.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b656ec5fad162dcf918b3fd8c696b8b48e129c3eaf4a68480bac03bfd90ccac
+size 14949
diff --git a/checkpoints/step_3780.pt b/checkpoints/step_3780.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ee50558209e058000e48153ed212d07006044fdc
--- /dev/null
+++ b/checkpoints/step_3780.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0208c3a1e97cfdf1b5a359302e72c83d4d854bff600a3969fdcf4416fb1527c2
+size 14949
diff --git a/checkpoints/step_3840.pt b/checkpoints/step_3840.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a76111be71b03f73b18b0b33458a7a517290a6ba
--- /dev/null
+++ b/checkpoints/step_3840.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e19f05d656414d3b1f81fc51eb158051a324616f7fc5bd549eee38960c89ad2
+size 14949
diff --git a/checkpoints/step_3900.pt b/checkpoints/step_3900.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3135c4063bbb3f0c27793921ff564b5bddf149da
--- /dev/null
+++ b/checkpoints/step_3900.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8676d0fd3322c40c154ad6eea341bce03e495592d0278dc5e6b639a668b17db4
+size 14949
diff --git a/checkpoints/step_3960.pt b/checkpoints/step_3960.pt
new file mode 100644
index 0000000000000000000000000000000000000000..bc9323d99035f683ada489d6ad952868192133b2
--- /dev/null
+++ b/checkpoints/step_3960.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12c3edd37678099617dc74251c34acbed50b16d1048b5cfe84b9e118b0c5abbc
+size 14949
diff --git a/checkpoints/step_4020.pt b/checkpoints/step_4020.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4ee07eeb27611b2e805242b85f74fe5eba00c38c
--- /dev/null
+++ b/checkpoints/step_4020.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01738c20da708cd642161cdf03fc6083ad07686bc708a311a514c4ffd7856237
+size 14949
diff --git a/checkpoints/step_4080.pt b/checkpoints/step_4080.pt
new file mode 100644
index 0000000000000000000000000000000000000000..62f8f656d3b360f97710d5726703ea3d01f628cb
--- /dev/null
+++ b/checkpoints/step_4080.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d36f570472258fa37aa8cfe892d230fe5360dd309026b24a718c3db97844bd7e
+size 14949
diff --git a/checkpoints/step_4140.pt b/checkpoints/step_4140.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8f63ee200663c9f8b213f6c969b537a56f4d1dd0
--- /dev/null
+++ b/checkpoints/step_4140.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a8ae20be4862861dea18cfe81e254b44997d7e5ee1f320931486a3c3e82f00e
+size 14949
diff --git a/checkpoints/step_420.pt b/checkpoints/step_420.pt
new file mode 100644
index 0000000000000000000000000000000000000000..57fc1255d26b57caac51cc2e09de801eb80c204b
--- /dev/null
+++ b/checkpoints/step_420.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f74d4aeba5fda5e8487a874d7d3b367b742ec39420d581fceefcada58cc22e2
+size 14933
diff --git a/checkpoints/step_4200.pt b/checkpoints/step_4200.pt
new file mode 100644
index 0000000000000000000000000000000000000000..460ef0fd45de9d2de6c6f369785b32e2386b309a
--- /dev/null
+++ b/checkpoints/step_4200.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08c7a3d0534aaafd8c158786bb91b3e12259aba31021ce71f9bdbf5d85e191d1
+size 14949
diff --git a/checkpoints/step_4260.pt b/checkpoints/step_4260.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a5f21345dcb182d93f89de97f94f76d8768034f8
--- /dev/null
+++ b/checkpoints/step_4260.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f63e957a5a212260cf6c5bfa8d523e3124a3f6ccc0dd831c99d32ad199ea610
+size 14949
diff --git a/checkpoints/step_4320.pt b/checkpoints/step_4320.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d86cf2571593aff44c1badff8d21a4c6b4d332cf
--- /dev/null
+++ b/checkpoints/step_4320.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5604f77c0891601cff855d33d03524bca502736305c015462787371a4c1ecaa0
+size 14949
diff --git a/checkpoints/step_4380.pt b/checkpoints/step_4380.pt
new file mode 100644
index 0000000000000000000000000000000000000000..50002eb3fe2a49104906a6137c10cb654521925a
--- /dev/null
+++ b/checkpoints/step_4380.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cc1c77965774aa2744803be91eecf185c3a520a9ad2aa8a6f520725a8662689
+size 14949
diff --git a/checkpoints/step_4440.pt b/checkpoints/step_4440.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e58ff3fbd1e134c6d20ba302011687aa668892dd
--- /dev/null
+++ b/checkpoints/step_4440.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ca865351c4e7679c989ee850fe2b9e5f94f3dab8f56cd200a42f0775c11de26
+size 14949
diff --git a/checkpoints/step_4500.pt b/checkpoints/step_4500.pt
new file mode 100644
index 0000000000000000000000000000000000000000..146114613c3cc959321ea25fff80a3794efc00a1
--- /dev/null
+++ b/checkpoints/step_4500.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ff48024f5b30361e0784d7a0e0db825d1849e491de048a6fbdfc8b131385826
+size 14949
diff --git a/checkpoints/step_4560.pt b/checkpoints/step_4560.pt
new file mode 100644
index 0000000000000000000000000000000000000000..98bb49f9b21b23b7e9f59684865f4052e393876b
--- /dev/null
+++ b/checkpoints/step_4560.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eac684b3e12d3bc63d228b8c5dc290212f944dc7a9fb1bde9ce44d3438fa9b6
+size 14949
diff --git a/checkpoints/step_4620.pt b/checkpoints/step_4620.pt
new file mode 100644
index 0000000000000000000000000000000000000000..142ef5964a602d6947d3e80484b034f431f61384
--- /dev/null
+++ b/checkpoints/step_4620.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90e5bb19baa263a5fb5e13279b6dda97b8306d0f7647eaf730e4ba7035eff5d9
+size 14949
diff --git a/checkpoints/step_4680.pt b/checkpoints/step_4680.pt
new file mode 100644
index 0000000000000000000000000000000000000000..cd7c9db419ab22c8442229d2945463cf322561d4
--- /dev/null
+++ b/checkpoints/step_4680.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66e367d805df8f4c22ca8d3aac321a9c0dd12823406b413663837fa6275a5d85
+size 14949
diff --git a/checkpoints/step_4740.pt b/checkpoints/step_4740.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1b2b3a92836843b6cf149fb84d10532af30f5f3c
--- /dev/null
+++ b/checkpoints/step_4740.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a29006ec5f8ef5716aff14dbb476aa9400d8cc98aff7c4220cce25f550817423
+size 14949
diff --git a/checkpoints/step_480.pt b/checkpoints/step_480.pt
new file mode 100644
index 0000000000000000000000000000000000000000..99a7746ac598b1b3bd46d2e5ee3d54a40cb1bb69
--- /dev/null
+++ b/checkpoints/step_480.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76e40aa73a9a485dcb476305d9ce465babc8c3aa2fcbb596a25f4d16207a748b
+size 14933
diff --git a/checkpoints/step_4800.pt b/checkpoints/step_4800.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d78617b6fa941aae17d92f645ae2c00feb677ff7
--- /dev/null
+++ b/checkpoints/step_4800.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a6549dff31de42ddbe8cf43a4a9f1d8fbcfca7d783430079035fc652bdc1555
+size 14949
diff --git a/checkpoints/step_4860.pt b/checkpoints/step_4860.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8b1aa524f42b5620e98cf763b0dd829f8dedf98b
--- /dev/null
+++ b/checkpoints/step_4860.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0f3a1bbd673aed95f46b4d8a25e2f5427ef08881b0a48e130d947752f8c3ce3
+size 14949
diff --git a/checkpoints/step_4920.pt b/checkpoints/step_4920.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c4db42db1de745317d2bade2f4eca9037fccd00c
--- /dev/null
+++ b/checkpoints/step_4920.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdc428af451736c015a1fed32687600bdc96c7aa3f57b8889639c7474ddc794a
+size 14949
diff --git a/checkpoints/step_4980.pt b/checkpoints/step_4980.pt
new file mode 100644
index 0000000000000000000000000000000000000000..446ffe5a15dfae736c85f5c59a7dae22298cef20
--- /dev/null
+++ b/checkpoints/step_4980.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c8b32350afe9f3b69639b2449760b3f33f1e09234bb09f32ebe48ea7284c170
+size 14949
diff --git a/checkpoints/step_5040.pt b/checkpoints/step_5040.pt
new file mode 100644
index 0000000000000000000000000000000000000000..24fe88be381c1e95899b3dee71ea794bba24ec50
--- /dev/null
+++ b/checkpoints/step_5040.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12dc3dff47b6c96bb7c68c1f924594488d7ad96202507d17e374309f27da8eac
+size 14949
diff --git a/checkpoints/step_5100.pt b/checkpoints/step_5100.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c354b05534e9f958bd20e22ba0d83513a6d75f6c
--- /dev/null
+++ b/checkpoints/step_5100.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02e3be5b942c5a1917e8f75963b0d49803779d57c8f69002c7929bfa127ff947
+size 14949
diff --git a/checkpoints/step_5160.pt b/checkpoints/step_5160.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e6fd1cc208a4ceeb49c993bbc09e1855aee99622
--- /dev/null
+++ b/checkpoints/step_5160.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b7057dba5814c384ac01615cc407031412d129881727845f5fa7c2bb52c1dae
+size 14949
diff --git a/checkpoints/step_5220.pt b/checkpoints/step_5220.pt
new file mode 100644
index 0000000000000000000000000000000000000000..262fdcd109b66d62d694e975b531ea8e21aeaedd
--- /dev/null
+++ b/checkpoints/step_5220.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5753450845642f82fad8c819702415019f4ce31ce6195c45434965f6b7b315ac
+size 14949
diff --git a/checkpoints/step_5280.pt b/checkpoints/step_5280.pt
new file mode 100644
index 0000000000000000000000000000000000000000..594483dad99ea14f760c2504289ad760d399c557
--- /dev/null
+++ b/checkpoints/step_5280.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7007785740661a99a59fb92961cd8e7bfa3d599fe50eec295891a386a51a2c87
+size 14949
diff --git a/checkpoints/step_5340.pt b/checkpoints/step_5340.pt
new file mode 100644
index 0000000000000000000000000000000000000000..af0f8b3c42953404ec623d4271f2c58f3e1be7fa
--- /dev/null
+++ b/checkpoints/step_5340.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49e4331e087774211e646f6e05dfee2df782935e94f1e2b9879c770b50094174
+size 14949
diff --git a/checkpoints/step_540.pt b/checkpoints/step_540.pt
new file mode 100644
index 0000000000000000000000000000000000000000..31665a77d560fe7911b2b62267000304f19fee95
--- /dev/null
+++ b/checkpoints/step_540.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66ed9355a0bb83ac5cd43145fb5c94a6208cf3455875c8b5f9143f8ab2e4847c
+size 14933
diff --git a/checkpoints/step_5400.pt b/checkpoints/step_5400.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3fcb3796920cbc9feb130c10f0119a79c3ef310a
--- /dev/null
+++ b/checkpoints/step_5400.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30225a8d2defc441d2320481a0050ccfe82b41c0a45daa55d5988423df6bfde1
+size 14949
diff --git a/checkpoints/step_5460.pt b/checkpoints/step_5460.pt
new file mode 100644
index 0000000000000000000000000000000000000000..15ba66cb1657e926aaa4c3df386b40583c45de66
--- /dev/null
+++ b/checkpoints/step_5460.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6115a53ab660848e121f93e1bb4b32889370761adc830c0d87acbb55056749d0
+size 14949
diff --git a/checkpoints/step_5520.pt b/checkpoints/step_5520.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e1c154f17ebb43964964bf7dd4e8b9277b0be13c
--- /dev/null
+++ b/checkpoints/step_5520.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edde5839f4809ab6166d6f3a06845cbd926a2ccd2471992183e0e0957194e089
+size 14949
diff --git a/checkpoints/step_5580.pt b/checkpoints/step_5580.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d19135d26c51cc3ac5361743bc5fdd6786e36325
--- /dev/null
+++ b/checkpoints/step_5580.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af4dc1d09f58fccd53c77ff4f9e0272d4eedfebe724d47c27fa0ff5fc82a57c5
+size 14949
diff --git a/checkpoints/step_5640.pt b/checkpoints/step_5640.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4c008bc86daf9458c5ad53403bffbcba541eadb3
--- /dev/null
+++ b/checkpoints/step_5640.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f79f5b6735a1922ae8d9f617771a01e9a75c896add800d1922e19cdc545d7f93
+size 14949
diff --git a/checkpoints/step_5700.pt b/checkpoints/step_5700.pt
new file mode 100644
index 0000000000000000000000000000000000000000..aef45fa15bafc43e8a417be028d697b51b36bf56
--- /dev/null
+++ b/checkpoints/step_5700.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f78a7ab55fe6fe1127adeb3fd7bdc452373da52b6e75968ec1b91bb6aebd315e
+size 14949
diff --git a/checkpoints/step_5760.pt b/checkpoints/step_5760.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6e3956e0ccb7204a38f378b6139eb992886b248e
--- /dev/null
+++ b/checkpoints/step_5760.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:536da26ed64155107bee7edaa40ce7a6e37a20ef38a3e6f397e893fbc780446a
+size 14949
diff --git a/checkpoints/step_5820.pt b/checkpoints/step_5820.pt
new file mode 100644
index 0000000000000000000000000000000000000000..66f193c00cf4d1659cfefc0f57d09fd2faa4e941
--- /dev/null
+++ b/checkpoints/step_5820.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66db31172b12bdc5eea4744aa19bfaf7aa79699579adb527128446998f676b50
+size 14949
diff --git a/checkpoints/step_5880.pt b/checkpoints/step_5880.pt
new file mode 100644
index 0000000000000000000000000000000000000000..082b7b4d14498d25e822864fb626eab9ac33cdd0
--- /dev/null
+++ b/checkpoints/step_5880.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf70386d71037fbc4e687e431d6669d28d01d3e7461c935f4bca524aa74dd719
+size 14949
diff --git a/checkpoints/step_5940.pt b/checkpoints/step_5940.pt
new file mode 100644
index 0000000000000000000000000000000000000000..5a566bbfbf5bbfb5f47ab8e60ec7b3f6fdbcd97a
--- /dev/null
+++ b/checkpoints/step_5940.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a705160b4f4acde5cf78f7d838696bc5d9ba780499dfa8eb4dc0fbc538f78274
+size 14949
diff --git a/checkpoints/step_60.pt b/checkpoints/step_60.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c2e771c441bcf77e2b1c1f7ae5a121c8c75dd402
--- /dev/null
+++ b/checkpoints/step_60.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9184eccbfdcb07e03423f6037f5c4c6685832ca3ac033db1026d1ce2634afa82
+size 14341
diff --git a/checkpoints/step_600.pt b/checkpoints/step_600.pt
new file mode 100644
index 0000000000000000000000000000000000000000..31b686d1cdd44ca1e7677876f16a1789ddb506ca
--- /dev/null
+++ b/checkpoints/step_600.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f14eb04c04272ac1370aef5db7941a14797c5ab71fb6215de73638dfc01d609d
+size 14933
diff --git a/checkpoints/step_6000.pt b/checkpoints/step_6000.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a97088ea451ffd5d144b1068626abe896501b2a6
--- /dev/null
+++ b/checkpoints/step_6000.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd634ed7d12ef43569a35d17c66814143bdb572fd3011a9ceeba9c52608f9d60
+size 14949
diff --git a/checkpoints/step_660.pt b/checkpoints/step_660.pt
new file mode 100644
index 0000000000000000000000000000000000000000..49145e630950faaa46df5772b11012d80e027eba
--- /dev/null
+++ b/checkpoints/step_660.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:274dcb83923c4c5f54645c0a3cf166d38855b231689f171683dc53512e7cc7e1
+size 14933
diff --git a/checkpoints/step_720.pt b/checkpoints/step_720.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e3c6336c188bd9b894626c130da9a4f936de34a1
--- /dev/null
+++ b/checkpoints/step_720.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38e5ff3672501a86c1791d4c622ca3e05bb17e0e043179595d37ada8675ddb81
+size 14933
diff --git a/checkpoints/step_780.pt b/checkpoints/step_780.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2c2f60f77fed83478b883a664321c421e0869937
--- /dev/null
+++ b/checkpoints/step_780.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d47d6c1a8ed6fc1716e5c14c89b6b321e9312770e8e2a11666c72f72bc94fa1
+size 14933
diff --git a/checkpoints/step_840.pt b/checkpoints/step_840.pt
new file mode 100644
index 0000000000000000000000000000000000000000..05a338955a3a224e723fc103c6d8218a82ef2087
--- /dev/null
+++ b/checkpoints/step_840.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef9461bc2bb372699e8fb949bbdecc449129afb6eb741c76e090572598686cb
+size 14933
diff --git a/checkpoints/step_900.pt b/checkpoints/step_900.pt
new file mode 100644
index 0000000000000000000000000000000000000000..fb93814715138b1a69fa3a06d69f662d1a455fdc
--- /dev/null
+++ b/checkpoints/step_900.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c514e976cba013435c2c389719f1bb2b8cf67b9b3ddb43bdb0a26a606d11bce1
+size 14933
diff --git a/checkpoints/step_960.pt b/checkpoints/step_960.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2006f5649680481d243af9ee812ed538f54e54c6
--- /dev/null
+++ b/checkpoints/step_960.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ca807de58de30b6b3c5092d92467de9edd8f8af8173f4f90044bebd31ec339c
+size 14933
diff --git a/config.yaml b/config.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..f7ea637f86c415acf08f13effc5bf49fc7b4d012
--- /dev/null
+++ b/config.yaml
@@ -0,0 +1,44 @@
+init:
+  init_type: mup
+  std_embed: 0.02
+  std_o: 0.01
+  std_qkv: 0.02
+loss:
+  label_smoothing: 0.0
+  type: next_token_ce
+model:
+  attn_scale: 0.35
+  attn_type: quadratic
+  d_model: 16
+  n_ctx: 16
+  n_head: 2
+  n_layers: 2
+  norm_places:
+  - pre_unembed
+  norm_type: none
+  rope_base: 10000
+  use_bias_qk: false
+  use_rmsnorm_qk: false
+  vocab_size: 16
+name: big_experiment
+seed: 42
+train:
+  batch_size: 128
+  betas:
+  - 0.9
+  - 0.95
+  checkpoint_every: 60
+  checkpoint_linear_count: 20
+  checkpoint_log_count: 0
+  checkpoint_schedule: interval
+  debug: true
+  dtype: bfloat16
+  eval_every: 100
+  grad_clip: 1.0
+  lr: 0.0003
+  lr_decay_frac: 1.0
+  max_steps: 6000
+  muon_lr: 0.03
+  use_muon: true
+  warmup_steps: 500
+  weight_decay: 0.01
diff --git a/empirical_global_heatmaps/empirical_global_raw.png b/empirical_global_heatmaps/empirical_global_raw.png
new file mode 100644
index 0000000000000000000000000000000000000000..37228e3b8ae1d2aa5c5ac098776d888ea5ae3653
--- /dev/null
+++ b/empirical_global_heatmaps/empirical_global_raw.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb16d848cd8e9af9e1b9a8b245a2988a6818a19304d27665aa1d315d3404873c
+size 101738
diff --git a/empirical_global_heatmaps/empirical_global_sim.png b/empirical_global_heatmaps/empirical_global_sim.png
new file mode 100644
index 0000000000000000000000000000000000000000..8efd33ca18d10b678e74b468b3c4026a7962c9f8
--- /dev/null
+++ b/empirical_global_heatmaps/empirical_global_sim.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e47085dac127e2ea672896a32f6c2ab7bc2f9e44a2fefc74feb335faf8127023
+size 111815
diff --git a/empirical_global_heatmaps/empirical_global_sim_summary.csv b/empirical_global_heatmaps/empirical_global_sim_summary.csv
new file mode 100644
index 0000000000000000000000000000000000000000..23a1772ee7b2c06ee628251698167d5f7a74e04e
--- /dev/null
+++ b/empirical_global_heatmaps/empirical_global_sim_summary.csv
@@ -0,0 +1,442 @@
+step_i,step_j,value
+0,0,1.0
+0,300,0.6940225109682399
+0,600,0.3217534347384291
+0,900,0.005333814009563617
+0,1200,-0.008603919079065133
+0,1500,0.006627747207791058
+0,1800,-0.0004374415325335143
+0,2100,-0.014022661188369265
+0,2400,-0.011151797810778182
+0,2700,-0.011144529903385115
+0,3000,0.0052197020704226405
+0,3300,0.005237933758221647
+0,3600,0.005207892000902545
+0,3900,0.002117996278150015
+0,4200,-0.004110605504600828
+0,4500,-0.008402177953579308
+0,4800,-0.0073879869594721806
+0,5100,-0.009154292814692266
+0,5400,-0.015693621769269955
+0,5700,0.000998877965762977
+0,6000,-0.0012662027129841137
+300,0,0.6940225109682399
+300,300,1.0
+300,600,0.7987263943594857
+300,900,0.01297969459559145
+300,1200,-0.04839941378978313
+300,1500,-0.056473729669992916
+300,1800,-0.06308651613347334
+300,2100,-0.07679619867485603
+300,2400,-0.07182001072014266
+300,2700,-0.05662657345449749
+300,3000,-0.03771986953514757
+300,3300,-0.03805895346591938
+300,3600,-0.0393083503325751
+300,3900,-0.041458353846192224
+300,4200,-0.026818324232188842
+300,4500,-0.03799780766630311
+300,4800,-0.0322485827989814
+300,5100,-0.028792489951732215
+300,5400,-0.042502167404277764
+300,5700,-0.023474999096397133
+300,6000,-0.027210959624087594
+600,0,0.3217534347384291
+600,300,0.7987263943594857
+600,600,1.0
+600,900,0.10456977054681325
+600,1200,-0.00017250811459045118
+600,1500,-0.021144648490525303
+600,1800,-0.020758756772964224
+600,2100,-0.038151074983991076
+600,2400,-0.025003219526070014
+600,2700,-0.01924515873951031
+600,3000,-0.003128694880824851
+600,3300,-0.007729804513493999
+600,3600,-0.010800068175206125
+600,3900,-0.002264650666204868
+600,4200,0.003980543382840997
+600,4500,-0.010063313458376836
+600,4800,-0.006912752067078996
+600,5100,0.004327411575043065
+600,5400,-0.008379912718226763
+600,5700,0.012755353675031257
+600,6000,0.0023455943405178343
+900,0,0.005333814009563617
+900,300,0.01297969459559145
+900,600,0.10456977054681325
+900,900,1.0
+900,1200,0.30874129271112116
+900,1500,0.25612166367722666
+900,1800,0.2630469883792215
+900,2100,0.2827262439279695
+900,2400,0.25795139414125884
+900,2700,0.2238196759538138
+900,3000,0.20764198575496073
+900,3300,0.1912283459790778
+900,3600,0.1782622304746056
+900,3900,0.16664156571920827
+900,4200,0.16033855074301157
+900,4500,0.16295158000098414
+900,4800,0.15597187903245469
+900,5100,0.15665420730717627
+900,5400,0.15419886933560675
+900,5700,0.1523373056779359
+900,6000,0.14850710267583825
+1200,0,-0.008603919079065133
+1200,300,-0.04839941378978313
+1200,600,-0.00017250811459045118
+1200,900,0.30874129271112116
+1200,1200,1.0
+1200,1500,0.6972555065551813
+1200,1800,0.6176368546484942
+1200,2100,0.5806105309142622
+1200,2400,0.4959197453555812
+1200,2700,0.37135270145196836
+1200,3000,0.3752869468125731
+1200,3300,0.33784289401273904
+1200,3600,0.31578004379641267
+1200,3900,0.2889532099557161
+1200,4200,0.28273400019278316
+1200,4500,0.2691043113999206
+1200,4800,0.26862381271732555
+1200,5100,0.26817838419415224
+1200,5400,0.26212426469425215
+1200,5700,0.25948517079901273
+1200,6000,0.2514251921151899
+1500,0,0.006627747207791058
+1500,300,-0.056473729669992916
+1500,600,-0.021144648490525303
+1500,900,0.25612166367722666
+1500,1200,0.6972555065551813
+1500,1500,1.0
+1500,1800,0.8622159104923277
+1500,2100,0.7837251962068084
+1500,2400,0.6598366363322931
+1500,2700,0.4873833653953313
+1500,3000,0.47399670151769197
+1500,3300,0.42104075075451397
+1500,3600,0.39516655830748515
+1500,3900,0.36785216818380817
+1500,4200,0.34658874236164683
+1500,4500,0.3374121609211241
+1500,4800,0.33895568126308784
+1500,5100,0.3285474751802855
+1500,5400,0.31783293693143283
+1500,5700,0.3299380686786487
+1500,6000,0.31936984786091693
+1800,0,-0.0004374415325335143
+1800,300,-0.06308651613347334
+1800,600,-0.020758756772964224
+1800,900,0.2630469883792215
+1800,1200,0.6176368546484942
+1800,1500,0.8622159104923277
+1800,1800,1.0
+1800,2100,0.8652202685189593
+1800,2400,0.7321402812504872
+1800,2700,0.5365188171299765
+1800,3000,0.5275534441747641
+1800,3300,0.4508748899228626
+1800,3600,0.42495162863126645
+1800,3900,0.39638694046263817
+1800,4200,0.3730261876105253
+1800,4500,0.3646006762466927
+1800,4800,0.36517884278237955
+1800,5100,0.3606054663988873
+1800,5400,0.3451437108796451
+1800,5700,0.35589509321606916
+1800,6000,0.33956276576551536
+2100,0,-0.014022661188369265
+2100,300,-0.07679619867485603
+2100,600,-0.038151074983991076
+2100,900,0.2827262439279695
+2100,1200,0.5806105309142622
+2100,1500,0.7837251962068084
+2100,1800,0.8652202685189593
+2100,2100,1.0
+2100,2400,0.8517103138500921
+2100,2700,0.6109139998476597
+2100,3000,0.5597219661774929
+2100,3300,0.47868848439846406
+2100,3600,0.44894637314743735
+2100,3900,0.4127788390093164
+2100,4200,0.38745457564613356
+2100,4500,0.38264133963524916
+2100,4800,0.3878312768611896
+2100,5100,0.3698671761917822
+2100,5400,0.35449876429364585
+2100,5700,0.36759072677412885
+2100,6000,0.3478872566342525
+2400,0,-0.011151797810778182
+2400,300,-0.07182001072014266
+2400,600,-0.025003219526070014
+2400,900,0.25795139414125884
+2400,1200,0.4959197453555812
+2400,1500,0.6598366363322931
+2400,1800,0.7321402812504872
+2400,2100,0.8517103138500921
+2400,2400,1.0
+2400,2700,0.7346717230156232
+2400,3000,0.6250061570010844
+2400,3300,0.5409850571046496
+2400,3600,0.4995611281583321
+2400,3900,0.4605534527522779
+2400,4200,0.43095958393849165
+2400,4500,0.4220627660634007
+2400,4800,0.4216129270985552
+2400,5100,0.4079784637930614
+2400,5400,0.3953996349290808
+2400,5700,0.40496707215062017
+2400,6000,0.39133276995572575
+2700,0,-0.011144529903385115
+2700,300,-0.05662657345449749
+2700,600,-0.01924515873951031
+2700,900,0.2238196759538138
+2700,1200,0.37135270145196836
+2700,1500,0.4873833653953313
+2700,1800,0.5365188171299765
+2700,2100,0.6109139998476597
+2700,2400,0.7346717230156232
+2700,2700,1.0
+2700,3000,0.7958588622968086
+2700,3300,0.6809083852994979
+2700,3600,0.5946758657961376
+2700,3900,0.5457699658927075
+2700,4200,0.5099847969262629
+2700,4500,0.5018942740203813
+2700,4800,0.49647011819216935
+2700,5100,0.48497754911797764
+2700,5400,0.4773374531298742
+2700,5700,0.4871827295341688
+2700,6000,0.4731555602159083
+3000,0,0.0052197020704226405
+3000,300,-0.03771986953514757
+3000,600,-0.003128694880824851
+3000,900,0.20764198575496073
+3000,1200,0.3752869468125731
+3000,1500,0.47399670151769197
+3000,1800,0.5275534441747641
+3000,2100,0.5597219661774929
+3000,2400,0.6250061570010844
+3000,2700,0.7958588622968086
+3000,3000,1.0
+3000,3300,0.8634888358039507
+3000,3600,0.7590784300994629
+3000,3900,0.7032465946090406
+3000,4200,0.6555795300197689
+3000,4500,0.6333087880369327
+3000,4800,0.6277293629015381
+3000,5100,0.616552213703373
+3000,5400,0.5884434848801534
+3000,5700,0.5990938276118499
+3000,6000,0.5756445545607052
+3300,0,0.005237933758221647
+3300,300,-0.03805895346591938
+3300,600,-0.007729804513493999
+3300,900,0.1912283459790778
+3300,1200,0.33784289401273904
+3300,1500,0.42104075075451397
+3300,1800,0.4508748899228626
+3300,2100,0.47868848439846406
+3300,2400,0.5409850571046496
+3300,2700,0.6809083852994979
+3300,3000,0.8634888358039507
+3300,3300,1.0
+3300,3600,0.8855567540615926
+3300,3900,0.8316164165430731
+3300,4200,0.7759509450461133
+3300,4500,0.7477971348324024
+3300,4800,0.7155058049577896
+3300,5100,0.709721674144009
+3300,5400,0.6859655012258002
+3300,5700,0.6836322227882338
+3300,6000,0.6691036351020759
+3600,0,0.005207892000902545
+3600,300,-0.0393083503325751
+3600,600,-0.010800068175206125
+3600,900,0.1782622304746056
+3600,1200,0.31578004379641267
+3600,1500,0.39516655830748515
+3600,1800,0.42495162863126645
+3600,2100,0.44894637314743735
+3600,2400,0.4995611281583321
+3600,2700,0.5946758657961376
+3600,3000,0.7590784300994629
+3600,3300,0.8855567540615926
+3600,3600,1.0
+3600,3900,0.9125585628480952
+3600,4200,0.8657856868977425
+3600,4500,0.8368121607874228
+3600,4800,0.8146676968357328
+3600,5100,0.7857051829991317
+3600,5400,0.7548508228596864
+3600,5700,0.7549790020430412
+3600,6000,0.7501136285866242
+3900,0,0.002117996278150015
+3900,300,-0.041458353846192224
+3900,600,-0.002264650666204868
+3900,900,0.16664156571920827
+3900,1200,0.2889532099557161
+3900,1500,0.36785216818380817
+3900,1800,0.39638694046263817
+3900,2100,0.4127788390093164
+3900,2400,0.4605534527522779
+3900,2700,0.5457699658927075
+3900,3000,0.7032465946090406
+3900,3300,0.8316164165430731
+3900,3600,0.9125585628480952
+3900,3900,1.0
+3900,4200,0.9170298345210696
+3900,4500,0.8900752592994292
+3900,4800,0.8502491923493983
+3900,5100,0.8387577056289179
+3900,5400,0.802245125019019
+3900,5700,0.7955649215520288
+3900,6000,0.7787796970641213
+4200,0,-0.004110605504600828
+4200,300,-0.026818324232188842
+4200,600,0.003980543382840997
+4200,900,0.16033855074301157
+4200,1200,0.28273400019278316
+4200,1500,0.34658874236164683
+4200,1800,0.3730261876105253
+4200,2100,0.38745457564613356
+4200,2400,0.43095958393849165
+4200,2700,0.5099847969262629
+4200,3000,0.6555795300197689
+4200,3300,0.7759509450461133
+4200,3600,0.8657856868977425
+4200,3900,0.9170298345210696
+4200,4200,1.0
+4200,4500,0.9261610152123264
+4200,4800,0.892847476898119
+4200,5100,0.880976747655371
+4200,5400,0.8601162452556256
+4200,5700,0.8431291617290139
+4200,6000,0.8296436058432634
+4500,0,-0.008402177953579308
+4500,300,-0.03799780766630311
+4500,600,-0.010063313458376836
+4500,900,0.16295158000098414
+4500,1200,0.2691043113999206
+4500,1500,0.3374121609211241
+4500,1800,0.3646006762466927
+4500,2100,0.38264133963524916
+4500,2400,0.4220627660634007
+4500,2700,0.5018942740203813
+4500,3000,0.6333087880369327
+4500,3300,0.7477971348324024
+4500,3600,0.8368121607874228
+4500,3900,0.8900752592994292
+4500,4200,0.9261610152123264
+4500,4500,1.0
+4500,4800,0.9350229097272709
+4500,5100,0.9256545995250737
+4500,5400,0.8963605393763684
+4500,5700,0.8899703319615357
+4500,6000,0.8711604848729894
+4800,0,-0.0073879869594721806
+4800,300,-0.0322485827989814
+4800,600,-0.006912752067078996
+4800,900,0.15597187903245469
+4800,1200,0.26862381271732555
+4800,1500,0.33895568126308784
+4800,1800,0.36517884278237955
+4800,2100,0.3878312768611896
+4800,2400,0.4216129270985552
+4800,2700,0.49647011819216935
+4800,3000,0.6277293629015381
+4800,3300,0.7155058049577896
+4800,3600,0.8146676968357328
+4800,3900,0.8502491923493983
+4800,4200,0.892847476898119
+4800,4500,0.9350229097272709
+4800,4800,1.0
+4800,5100,0.9264234411754766
+4800,5400,0.9029378851467372
+4800,5700,0.9042537469090977
+4800,6000,0.8755917664108668
+5100,0,-0.009154292814692266
+5100,300,-0.028792489951732215
+5100,600,0.004327411575043065
+5100,900,0.15665420730717627
+5100,1200,0.26817838419415224
+5100,1500,0.3285474751802855
+5100,1800,0.3606054663988873
+5100,2100,0.3698671761917822
+5100,2400,0.4079784637930614
+5100,2700,0.48497754911797764
+5100,3000,0.616552213703373
+5100,3300,0.709721674144009
+5100,3600,0.7857051829991317
+5100,3900,0.8387577056289179
+5100,4200,0.880976747655371
+5100,4500,0.9256545995250737
+5100,4800,0.9264234411754766
+5100,5100,1.0
+5100,5400,0.9072712210670539
+5100,5700,0.922266664334384
+5100,6000,0.8973589091454117
+5400,0,-0.015693621769269955
+5400,300,-0.042502167404277764
+5400,600,-0.008379912718226763
+5400,900,0.15419886933560675
+5400,1200,0.26212426469425215
+5400,1500,0.31783293693143283
+5400,1800,0.3451437108796451
+5400,2100,0.35449876429364585
+5400,2400,0.3953996349290808
+5400,2700,0.4773374531298742
+5400,3000,0.5884434848801534
+5400,3300,0.6859655012258002
+5400,3600,0.7548508228596864
+5400,3900,0.802245125019019
+5400,4200,0.8601162452556256
+5400,4500,0.8963605393763684
+5400,4800,0.9029378851467372
+5400,5100,0.9072712210670539
+5400,5400,1.0
+5400,5700,0.9149636197820636
+5400,6000,0.8921942916674346
+5700,0,0.000998877965762977
+5700,300,-0.023474999096397133
+5700,600,0.012755353675031257
+5700,900,0.1523373056779359
+5700,1200,0.25948517079901273
+5700,1500,0.3299380686786487
+5700,1800,0.35589509321606916
+5700,2100,0.36759072677412885
+5700,2400,0.40496707215062017
+5700,2700,0.4871827295341688
+5700,3000,0.5990938276118499
+5700,3300,0.6836322227882338
+5700,3600,0.7549790020430412
+5700,3900,0.7955649215520288
+5700,4200,0.8431291617290139
+5700,4500,0.8899703319615357
+5700,4800,0.9042537469090977
+5700,5100,0.922266664334384
+5700,5400,0.9149636197820636
+5700,5700,1.0
+5700,6000,0.932180524242329
+6000,0,-0.0012662027129841137
+6000,300,-0.027210959624087594
+6000,600,0.0023455943405178343
+6000,900,0.14850710267583825
+6000,1200,0.2514251921151899
+6000,1500,0.31936984786091693
+6000,1800,0.33956276576551536
+6000,2100,0.3478872566342525
+6000,2400,0.39133276995572575
+6000,2700,0.4731555602159083
+6000,3000,0.5756445545607052
+6000,3300,0.6691036351020759
+6000,3600,0.7501136285866242
+6000,3900,0.7787796970641213
+6000,4200,0.8296436058432634
+6000,4500,0.8711604848729894
+6000,4800,0.8755917664108668
+6000,5100,0.8973589091454117
+6000,5400,0.8921942916674346
+6000,5700,0.932180524242329
+6000,6000,1.0
diff --git a/empirical_global_sims.npz b/empirical_global_sims.npz
new file mode 100644
index 0000000000000000000000000000000000000000..6d955dd4df46dc3eee9d27f840f89f1db561de12
--- /dev/null
+++ b/empirical_global_sims.npz
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f80242bab3a2f0cefdcbe575fdd8e649d1627373654f9c1dc411a5db40cda91
+size 9806
diff --git a/empirical_l1_head_heatmaps/local_all_l1_heads.png b/empirical_l1_head_heatmaps/local_all_l1_heads.png
new file mode 100644
index 0000000000000000000000000000000000000000..8527eb6d908d7337397ccade31db3fada4b186e9
Binary files /dev/null and b/empirical_l1_head_heatmaps/local_all_l1_heads.png differ
diff --git a/empirical_l1_head_heatmaps/local_head0_vs_head0.png b/empirical_l1_head_heatmaps/local_head0_vs_head0.png
new file mode 100644
index 0000000000000000000000000000000000000000..393b88d6768540a1799b3e127e1122424419d1e3
--- /dev/null
+++ b/empirical_l1_head_heatmaps/local_head0_vs_head0.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcf97d5f33331572ccbd438b6aed0a802b20e5d645e20b794a9a8a24eac92b0c
+size 117560
diff --git a/empirical_l1_head_heatmaps/local_head0_vs_head1.png b/empirical_l1_head_heatmaps/local_head0_vs_head1.png
new file mode 100644
index 0000000000000000000000000000000000000000..fe706216f5813f584ae5623dadf84f2256ff410a
--- /dev/null
+++ b/empirical_l1_head_heatmaps/local_head0_vs_head1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74d23e8fdc28cfe593ddb0b7f681503b2bc67231c0b86f045d2006e5b0c42543
+size 116637
diff --git a/empirical_l1_head_heatmaps/local_head1_vs_head0.png b/empirical_l1_head_heatmaps/local_head1_vs_head0.png
new file mode 100644
index 0000000000000000000000000000000000000000..9c7682fbd0253c71c8a3302844887183eaa03d47
--- /dev/null
+++ b/empirical_l1_head_heatmaps/local_head1_vs_head0.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79fe22e247dd3f6ac5adeff3a1ab397bc9b7c460b1f96eefca211604b5aff1eb
+size 116621
diff --git a/empirical_l1_head_heatmaps/local_head1_vs_head1.png b/empirical_l1_head_heatmaps/local_head1_vs_head1.png
new file mode 100644
index 0000000000000000000000000000000000000000..eda94af18c9df0fbea2fb8331c20726ffa59a644
--- /dev/null
+++ b/empirical_l1_head_heatmaps/local_head1_vs_head1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bdb6718efea9e8f4cf13e760f73ad397e0d266bdd8d939c31bc93981db77f091
+size 116629
diff --git a/empirical_l1_head_heatmaps/local_summary.csv b/empirical_l1_head_heatmaps/local_summary.csv
new file mode 100644
index 0000000000000000000000000000000000000000..a2c8ff701a66a2c3c2e660c4e1e28d51b6d04415
--- /dev/null
+++ b/empirical_l1_head_heatmaps/local_summary.csv
@@ -0,0 +1,5 @@
+source_head,target_head,mean,min,max,finite_count
+head0,head0,0.47277070358068735,-0.12333237722941813,1.0000000000000095,441
+head0,head1,-0.041945918574264315,-0.24331884374818488,0.14837353359183098,441
+head1,head0,-0.04194591857426432,-0.24331884374818488,0.14837353359183098,441
+head1,head1,0.5089063087898688,-0.057515667712945036,0.9999999999999867,441
diff --git a/empirical_l1_head_sims.npz b/empirical_l1_head_sims.npz
new file mode 100644
index 0000000000000000000000000000000000000000..ad34ab4a1ee657a65dde757a872bf17f3a71c178
--- /dev/null
+++ b/empirical_l1_head_sims.npz
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e23d7a9706fd4c271a3f0c45f0411d57c4728bb01718fdf9475288036e548bd1
+size 31561
diff --git a/empirical_layer_block_heatmaps/local_all_layer_blocks.png b/empirical_layer_block_heatmaps/local_all_layer_blocks.png
new file mode 100644
index 0000000000000000000000000000000000000000..26c78051f21c7eb69fd7c4957c6cfbd6fb8e4762
Binary files /dev/null and b/empirical_layer_block_heatmaps/local_all_layer_blocks.png differ
diff --git a/empirical_layer_block_heatmaps/local_direct_vs_direct.png b/empirical_layer_block_heatmaps/local_direct_vs_direct.png
new file mode 100644
index 0000000000000000000000000000000000000000..43f12909d4d7e9449a2d3a850425895220cc1f92
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_direct_vs_direct.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b37ec5dd787424c0f21bece761cdc2d98b52bb31702d8d9fe71b916c4940347b
+size 114699
diff --git a/empirical_layer_block_heatmaps/local_direct_vs_layer1.png b/empirical_layer_block_heatmaps/local_direct_vs_layer1.png
new file mode 100644
index 0000000000000000000000000000000000000000..5fd451c3e6de11ac49417c72cd42ce0ab95cfe1a
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_direct_vs_layer1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d94f271e484866c9ef709f28cd149eb13ce25cfac4b8a2390c60f14e15d519f
+size 114773
diff --git a/empirical_layer_block_heatmaps/local_direct_vs_layer2.png b/empirical_layer_block_heatmaps/local_direct_vs_layer2.png
new file mode 100644
index 0000000000000000000000000000000000000000..911fb9f9e6d0f7bcb690dd3b2f5fe9b95a47b0eb
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_direct_vs_layer2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4b6792868f75c499700fc64636ad1d91db5be624e9ce61ec757ce3783a1288a
+size 114570
diff --git a/empirical_layer_block_heatmaps/local_layer1_vs_direct.png b/empirical_layer_block_heatmaps/local_layer1_vs_direct.png
new file mode 100644
index 0000000000000000000000000000000000000000..a71cdcbd55e4be05b24685f5018efe11b5b019e3
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_layer1_vs_direct.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ad86b9d64bba340013f095dee9ffa5cdee588ac4521e6a570238ed7bbf81e04
+size 114691
diff --git a/empirical_layer_block_heatmaps/local_layer1_vs_layer1.png b/empirical_layer_block_heatmaps/local_layer1_vs_layer1.png
new file mode 100644
index 0000000000000000000000000000000000000000..39146ab3cfab02547c4c48d4644bdc115c0e12c4
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_layer1_vs_layer1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f19e8832737dbd8c56a572760bb3fc17d077880cafe13794d7706cfa86ba492
+size 114736
diff --git a/empirical_layer_block_heatmaps/local_layer1_vs_layer2.png b/empirical_layer_block_heatmaps/local_layer1_vs_layer2.png
new file mode 100644
index 0000000000000000000000000000000000000000..604ed99dc9c54854008f083fe0bfb00cb16e2117
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_layer1_vs_layer2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c29cc1a83ad3771e223e1c2140dcd3e94245ebaad2425eea9e27602789e883b
+size 114505
diff --git a/empirical_layer_block_heatmaps/local_layer2_vs_direct.png b/empirical_layer_block_heatmaps/local_layer2_vs_direct.png
new file mode 100644
index 0000000000000000000000000000000000000000..9ae9e2ed9dd029d2f3ebbd4313548bf74ca168b6
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_layer2_vs_direct.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd5ac242e503b197b832ce46ab46a5fd2696a2f296a71b65ea870f52b543385b
+size 114421
diff --git a/empirical_layer_block_heatmaps/local_layer2_vs_layer1.png b/empirical_layer_block_heatmaps/local_layer2_vs_layer1.png
new file mode 100644
index 0000000000000000000000000000000000000000..2143fda2dfbb89b2f869b5861e4bd7f9516628f4
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_layer2_vs_layer1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbebc8df770e1dedb39be26968daf96c02f3bc67ca8949e34aae8b4bb08d9325
+size 114476
diff --git a/empirical_layer_block_heatmaps/local_layer2_vs_layer2.png b/empirical_layer_block_heatmaps/local_layer2_vs_layer2.png
new file mode 100644
index 0000000000000000000000000000000000000000..96458612e5b4766156edbfc9d181b00cafed8830
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_layer2_vs_layer2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bd41836475a81784b77791303c171c1678dc8d77980fe4c2d0fae8f54dba903
+size 115155
diff --git a/empirical_layer_block_heatmaps/local_summary.csv b/empirical_layer_block_heatmaps/local_summary.csv
new file mode 100644
index 0000000000000000000000000000000000000000..1c32d833cc7086a8727ab869c1491270fd11357f
--- /dev/null
+++ b/empirical_layer_block_heatmaps/local_summary.csv
@@ -0,0 +1,10 @@
+source_group,target_group,mean,min,max,finite_count
+direct,direct,0.8716219756224064,0.20099064745859457,1.0000000000009321,441
+direct,layer1,-0.24112047733466715,-0.44623793674888024,0.140515109767336,441
+direct,layer2,-0.040179950703186036,-0.09231277545250804,0.05767510823693309,441
+layer1,direct,-0.24112047733466713,-0.44623793674888024,0.140515109767336,441
+layer1,layer1,0.518193386615448,-0.17454206319577295,0.9999999999999643,441
+layer1,layer2,0.06909030593165333,-0.11711985292053663,0.29736031211500036,441
+layer2,direct,-0.040179950703186036,-0.09231277545250804,0.05767510823693309,441
+layer2,layer1,0.06909030593165333,-0.11711985292053663,0.29736031211500036,441
+layer2,layer2,0.4352659020626333,-0.025770439021490725,1.000000000000065,441
diff --git a/empirical_layer_block_sims.npz b/empirical_layer_block_sims.npz
new file mode 100644
index 0000000000000000000000000000000000000000..9d509cc046ca9cf0bdf1b328cd34f8e97655bce3
--- /dev/null
+++ b/empirical_layer_block_sims.npz
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a18aa9025c67f83b731a13b3d61929af3ee47837cce1a9848a786073d9486641
+size 66912
diff --git a/final.pt b/final.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6508752b647d9be90ead1d8993c723a73830fde7
--- /dev/null
+++ b/final.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d4d6072c46703989f750107feb3341661d1f8ea2023276050d4afc6f374a00b
+size 14245
diff --git a/global_tn_heatmaps/global_tn_raw.png b/global_tn_heatmaps/global_tn_raw.png
new file mode 100644
index 0000000000000000000000000000000000000000..51d2112cefc8f30283d42eefd702727ee8a24950
Binary files /dev/null and b/global_tn_heatmaps/global_tn_raw.png differ
diff --git a/global_tn_heatmaps/global_tn_sim.png b/global_tn_heatmaps/global_tn_sim.png
new file mode 100644
index 0000000000000000000000000000000000000000..8d65b625254a6bb34d6844ade664010607540ffc
--- /dev/null
+++ b/global_tn_heatmaps/global_tn_sim.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bdb5305669bb35f6459553bfee7197d100280d98ded229c1dad95b119e9ae7c0
+size 109870
diff --git a/global_tn_heatmaps/global_tn_sim_summary.csv b/global_tn_heatmaps/global_tn_sim_summary.csv
new file mode 100644
index 0000000000000000000000000000000000000000..b1b01ed4b7df8050536cd8b13d19c4745f54f5dc
--- /dev/null
+++ b/global_tn_heatmaps/global_tn_sim_summary.csv
@@ -0,0 +1,442 @@
+step_i,step_j,value
+0,0,1.0
+0,300,0.6929716894115113
+0,600,-0.005317309841853853
+0,900,-0.011940414351313426
+0,1200,-0.012302135773632013
+0,1500,-0.009682702585007174
+0,1800,-0.00346102938280063
+0,2100,-0.006085858475656059
+0,2400,-0.012448000792716624
+0,2700,-0.010846114741838792
+0,3000,-0.00821590343515121
+0,3300,0.0015945981819391886
+0,3600,0.005367757678793949
+0,3900,0.0032182296959179005
+0,4200,0.011958569609888848
+0,4500,0.013881431626353708
+0,4800,0.01927226583015365
+0,5100,0.019299088913576
+0,5400,0.029669840462992314
+0,5700,0.03480549167615385
+0,6000,0.03316768760944894
+300,0,0.6929716894115112
+300,300,1.0
+300,600,0.018908491731851503
+300,900,-0.042092369298159994
+300,1200,-0.05038610643628463
+300,1500,-0.04806104731855837
+300,1800,-0.05632637052445746
+300,2100,-0.06308819249572917
+300,2400,-0.06180418710426462
+300,2700,-0.0453904181380134
+300,3000,-0.01272761747247402
+300,3300,0.0062189282392363775
+300,3600,0.008024460953206355
+300,3900,0.004287831197171499
+300,4200,0.019310532303868515
+300,4500,0.022400314102116212
+300,4800,0.035050076879281994
+300,5100,0.03475953778395217
+300,5400,0.05315491076360646
+300,5700,0.05733896843126829
+300,6000,0.05067162758548724
+600,0,-0.005317309841853852
+600,300,0.018908491731851506
+600,600,1.0
+600,900,-0.02228515434044147
+600,1200,-0.019301270550470987
+600,1500,-0.021401234303883403
+600,1800,-0.027501321796811876
+600,2100,-0.029833124874620053
+600,2400,-0.013767166539862628
+600,2700,-0.009835036584172303
+600,3000,-0.0035769882700495154
+600,3300,-0.01795962195974412
+600,3600,-0.00919478310334971
+600,3900,-0.009351850024749534
+600,4200,0.00017627360164308942
+600,4500,-0.0002085116783209785
+600,4800,0.003991774911445436
+600,5100,0.006617719976402866
+600,5400,0.016895484119096218
+600,5700,0.014843177665581707
+600,6000,0.0008078991756123058
+900,0,-0.011940414351313424
+900,300,-0.04209236929816
+900,600,-0.02228515434044147
+900,900,1.0
+900,1200,0.4055673738471332
+900,1500,0.3391361034384757
+900,1800,0.41823576969801324
+900,2100,0.3324697843363167
+900,2400,0.2810105133314378
+900,2700,0.23295857765207895
+900,3000,0.08016305443049614
+900,3300,-0.0007313842090760091
+900,3600,-0.016073523647394235
+900,3900,-0.030867904053815457
+900,4200,-0.03590284649460863
+900,4500,-0.05005148360545049
+900,4800,-0.08656486331268459
+900,5100,-0.1304080353863542
+900,5400,-0.13925384373333952
+900,5700,-0.13107623508335464
+900,6000,-0.1521223349560361
+1200,0,-0.012302135773632011
+1200,300,-0.050386106436284626
+1200,600,-0.01930127055047099
+1200,900,0.4055673738471332
+1200,1200,1.0
+1200,1500,0.830899372450385
+1200,1800,0.7722634122150874
+1200,2100,0.7530509839218088
+1200,2400,0.6502216799474049
+1200,2700,0.5221485338503843
+1200,3000,0.20685118445789813
+1200,3300,0.03530053415387606
+1200,3600,0.01593150857923303
+1200,3900,-0.022067042760133264
+1200,4200,-0.04785074965770543
+1200,4500,-0.04196966928958673
+1200,4800,-0.07725839178246903
+1200,5100,-0.13657541363235606
+1200,5400,-0.0860275284086592
+1200,5700,-0.10229028220264752
+1200,6000,-0.1440269781233677
+1500,0,-0.009682702585007176
+1500,300,-0.04806104731855839
+1500,600,-0.021401234303883403
+1500,900,0.3391361034384758
+1500,1200,0.8308993724503851
+1500,1500,1.0
+1500,1800,0.8244328523026531
+1500,2100,0.7656803067759097
+1500,2400,0.7284656922749606
+1500,2700,0.6066583471912066
+1500,3000,0.17065928276804038
+1500,3300,-0.040154923808805865
+1500,3600,0.008786345248838407
+1500,3900,-0.021430204105599433
+1500,4200,-0.042819692162152534
+1500,4500,-0.030188351563927254
+1500,4800,-0.06375959779129031
+1500,5100,-0.10529287790211216
+1500,5400,-0.07703304980281887
+1500,5700,-0.08513461326062809
+1500,6000,-0.1077136779543075
+1800,0,-0.0034610293828006293
+1800,300,-0.05632637052445747
+1800,600,-0.027501321796811872
+1800,900,0.41823576969801324
+1800,1200,0.7722634122150873
+1800,1500,0.8244328523026531
+1800,1800,1.0
+1800,2100,0.8255398047530393
+1800,2400,0.686697839017727
+1800,2700,0.5544633298578384
+1800,3000,0.1640668230248292
+1800,3300,-0.01756233316120226
+1800,3600,-0.015247369755831453
+1800,3900,-0.03862345519743803
+1800,4200,-0.0512706806564454
+1800,4500,-0.0472754646649555
+1800,4800,-0.0894562104562735
+1800,5100,-0.15487343278057636
+1800,5400,-0.13206776966292705
+1800,5700,-0.14893344047564422
+1800,6000,-0.16176859181848383
+2100,0,-0.006085858475656058
+2100,300,-0.06308819249572914
+2100,600,-0.029833124874620053
+2100,900,0.3324697843363167
+2100,1200,0.753050983921809
+2100,1500,0.7656803067759096
+2100,1800,0.8255398047530392
+2100,2100,1.0
+2100,2400,0.8047263483934827
+2100,2700,0.622109845504098
+2100,3000,0.2427854192566502
+2100,3300,0.03250113260704923
+2100,3600,0.004428590777709057
+2100,3900,-0.030641873441628654
+2100,4200,-0.04838854165114199
+2100,4500,-0.039918705623530726
+2100,4800,-0.0790570426130217
+2100,5100,-0.139104010798122
+2100,5400,-0.12172490083729481
+2100,5700,-0.13942340489118088
+2100,6000,-0.16715588251437632
+2400,0,-0.012448000792716622
+2400,300,-0.06180418710426465
+2400,600,-0.01376716653986263
+2400,900,0.2810105133314378
+2400,1200,0.6502216799474049
+2400,1500,0.7284656922749605
+2400,1800,0.6866978390177267
+2400,2100,0.8047263483934827
+2400,2400,1.0
+2400,2700,0.7897515583689112
+2400,3000,0.28357878642074824
+2400,3300,0.019699118047104138
+2400,3600,-0.004423854651887478
+2400,3900,-0.04062406799099273
+2400,4200,-0.048845374156307186
+2400,4500,-0.03829241044908618
+2400,4800,-0.08143950833195715
+2400,5100,-0.12933823420157411
+2400,5400,-0.11442734858934908
+2400,5700,-0.12115508875754594
+2400,6000,-0.1243799998772511
+2700,0,-0.010846114741838789
+2700,300,-0.04539041813801339
+2700,600,-0.009835036584172304
+2700,900,0.23295857765207895
+2700,1200,0.5221485338503843
+2700,1500,0.6066583471912066
+2700,1800,0.5544633298578384
+2700,2100,0.622109845504098
+2700,2400,0.7897515583689111
+2700,2700,1.0
+2700,3000,0.6188205867239283
+2700,3300,0.2750732383036427
+2700,3600,0.14180637382444525
+2700,3900,0.05219848556025994
+2700,4200,-0.00587582217761133
+2700,4500,-0.013294475576293128
+2700,4800,-0.05860001434695425
+2700,5100,-0.11758028138142032
+2700,5400,-0.07702353959394292
+2700,5700,-0.09418100159972093
+2700,6000,-0.1642665215428673
+3000,0,-0.008215903435151207
+3000,300,-0.012727617472474018
+3000,600,-0.003576988270049516
+3000,900,0.08016305443049614
+3000,1200,0.20685118445789813
+3000,1500,0.17065928276804035
+3000,1800,0.1640668230248292
+3000,2100,0.24278541925665026
+3000,2400,0.28357878642074824
+3000,2700,0.6188205867239285
+3000,3000,1.0
+3000,3300,0.7817875403834028
+3000,3600,0.5401616922827303
+3000,3900,0.3113904815958277
+3000,4200,0.17423140051758768
+3000,4500,0.10331726016751562
+3000,4800,0.09843459549023349
+3000,5100,0.12726281306624893
+3000,5400,0.05098513945851043
+3000,5700,0.04661470053289983
+3000,6000,-0.020162000341875323
+3300,0,0.001594598181939189
+3300,300,0.006218928239236377
+3300,600,-0.017959621959744123
+3300,900,-0.0007313842090760091
+3300,1200,0.03530053415387606
+3300,1500,-0.040154923808805865
+3300,1800,-0.01756233316120226
+3300,2100,0.03250113260704923
+3300,2400,0.019699118047104138
+3300,2700,0.2750732383036427
+3300,3000,0.7817875403834028
+3300,3300,1.0
+3300,3600,0.7593352726882207
+3300,3900,0.48633277473261355
+3300,4200,0.3089200628622702
+3300,4500,0.19628570870810355
+3300,4800,0.22638805421030134
+3300,5100,0.29604811165601347
+3300,5400,0.14981860540407807
+3300,5700,0.1561041187005239
+3300,6000,0.1735883948153407
+3600,0,0.00536775767879395
+3600,300,0.008024460953206358
+3600,600,-0.009194783103349709
+3600,900,-0.016073523647394238
+3600,1200,0.01593150857923303
+3600,1500,0.00878634524883841
+3600,1800,-0.01524736975583145
+3600,2100,0.0044285907777090565
+3600,2400,-0.004423854651887477
+3600,2700,0.14180637382444528
+3600,3000,0.5401616922827304
+3600,3300,0.7593352726882205
+3600,3600,1.0
+3600,3900,0.8078868439003566
+3600,4200,0.5682351328984674
+3600,4500,0.41163225422125715
+3600,4800,0.4267901673745769
+3600,5100,0.5195080535686738
+3600,5400,0.2756571633242121
+3600,5700,0.2751791234205781
+3600,6000,0.2988859380399491
+3900,0,0.0032182296959179023
+3900,300,0.0042878311971715
+3900,600,-0.009351850024749534
+3900,900,-0.030867904053815457
+3900,1200,-0.022067042760133267
+3900,1500,-0.021430204105599437
+3900,1800,-0.03862345519743803
+3900,2100,-0.030641873441628647
+3900,2400,-0.04062406799099273
+3900,2700,0.052198485560259945
+3900,3000,0.3113904815958278
+3900,3300,0.4863327747326135
+3900,3600,0.8078868439003565
+3900,3900,1.0
+3900,4200,0.8360669870199895
+3900,4500,0.725562943218162
+3900,4800,0.6741195603277135
+3900,5100,0.6329951765498956
+3900,5400,0.4245619467690822
+3900,5700,0.3456147876316629
+3900,6000,0.33953052421152685
+4200,0,0.011958569609888848
+4200,300,0.019310532303868522
+4200,600,0.00017627360164308945
+4200,900,-0.035902846494608634
+4200,1200,-0.047850749657705416
+4200,1500,-0.04281969216215255
+4200,1800,-0.05127068065644539
+4200,2100,-0.04838854165114199
+4200,2400,-0.04884537415630717
+4200,2700,-0.005875822177611331
+4200,3000,0.1742314005175877
+4200,3300,0.3089200628622702
+4200,3600,0.5682351328984673
+4200,3900,0.8360669870199895
+4200,4200,1.0
+4200,4500,0.8836087576319175
+4200,4800,0.8220975506631744
+4200,5100,0.6743762143620105
+4200,5400,0.613436040180404
+4200,5700,0.49962728889889296
+4200,6000,0.4165766616813595
+4500,0,0.013881431626353708
+4500,300,0.022400314102116216
+4500,600,-0.00020851167832097853
+4500,900,-0.05005148360545049
+4500,1200,-0.041969669289586724
+4500,1500,-0.03018835156392725
+4500,1800,-0.047275464664955504
+4500,2100,-0.03991870562353073
+4500,2400,-0.03829241044908617
+4500,2700,-0.013294475576293126
+4500,3000,0.10331726016751562
+4500,3300,0.19628570870810355
+4500,3600,0.4116322542212571
+4500,3900,0.7255629432181622
+4500,4200,0.8836087576319175
+4500,4500,1.0
+4500,4800,0.9201637235259682
+4500,5100,0.6725190799044697
+4500,5400,0.6843528922571313
+4500,5700,0.5224886975001574
+4500,6000,0.3865211915150546
+4800,0,0.01927226583015365
+4800,300,0.03505007687928199
+4800,600,0.003991774911445436
+4800,900,-0.08656486331268459
+4800,1200,-0.07725839178246902
+4800,1500,-0.06375959779129031
+4800,1800,-0.0894562104562735
+4800,2100,-0.07905704261302171
+4800,2400,-0.08143950833195715
+4800,2700,-0.058600014346954256
+4800,3000,0.0984345954902335
+4800,3300,0.22638805421030134
+4800,3600,0.4267901673745769
+4800,3900,0.6741195603277135
+4800,4200,0.8220975506631745
+4800,4500,0.9201637235259685
+4800,4800,1.0
+4800,5100,0.832520025677927
+4800,5400,0.7894127603455451
+4800,5700,0.6458409612636306
+4800,6000,0.5542893260294052
+5100,0,0.01929908891357599
+5100,300,0.03475953778395217
+5100,600,0.006617719976402867
+5100,900,-0.13040803538635423
+5100,1200,-0.13657541363235606
+5100,1500,-0.10529287790211214
+5100,1800,-0.15487343278057636
+5100,2100,-0.139104010798122
+5100,2400,-0.12933823420157414
+5100,2700,-0.11758028138142033
+5100,3000,0.12726281306624893
+5100,3300,0.2960481116560134
+5100,3600,0.5195080535686738
+5100,3900,0.6329951765498956
+5100,4200,0.6743762143620105
+5100,4500,0.6725190799044697
+5100,4800,0.8325200256779272
+5100,5100,1.0
+5100,5400,0.6492807932226976
+5100,5700,0.6043295729599548
+5100,6000,0.6973666901322939
+5400,0,0.029669840462992314
+5400,300,0.053154910763606456
+5400,600,0.016895484119096218
+5400,900,-0.13925384373333954
+5400,1200,-0.08602752840865918
+5400,1500,-0.07703304980281887
+5400,1800,-0.13206776966292705
+5400,2100,-0.12172490083729481
+5400,2400,-0.11442734858934907
+5400,2700,-0.07702353959394291
+5400,3000,0.05098513945851043
+5400,3300,0.1498186054040781
+5400,3600,0.27565716332421214
+5400,3900,0.4245619467690822
+5400,4200,0.6134360401804039
+5400,4500,0.6843528922571314
+5400,4800,0.789412760345545
+5400,5100,0.6492807932226977
+5400,5400,1.0
+5400,5700,0.9126591536592932
+5400,6000,0.630558391351182
+5700,0,0.03480549167615385
+5700,300,0.057338968431268296
+5700,600,0.014843177665581709
+5700,900,-0.13107623508335464
+5700,1200,-0.10229028220264755
+5700,1500,-0.08513461326062807
+5700,1800,-0.14893344047564422
+5700,2100,-0.13942340489118088
+5700,2400,-0.12115508875754591
+5700,2700,-0.09418100159972093
+5700,3000,0.04661470053289982
+5700,3300,0.15610411870052393
+5700,3600,0.27517912342057815
+5700,3900,0.3456147876316629
+5700,4200,0.49962728889889296
+5700,4500,0.5224886975001571
+5700,4800,0.6458409612636306
+5700,5100,0.6043295729599548
+5700,5400,0.9126591536592934
+5700,5700,1.0
+5700,6000,0.7287276302428333
+6000,0,0.03316768760944894
+6000,300,0.05067162758548723
+6000,600,0.0008078991756123058
+6000,900,-0.1521223349560361
+6000,1200,-0.1440269781233677
+6000,1500,-0.1077136779543075
+6000,1800,-0.16176859181848383
+6000,2100,-0.16715588251437635
+6000,2400,-0.1243799998772511
+6000,2700,-0.16426652154286728
+6000,3000,-0.020162000341875323
+6000,3300,0.1735883948153407
+6000,3600,0.2988859380399491
+6000,3900,0.3395305242115268
+6000,4200,0.4165766616813595
+6000,4500,0.3865211915150546
+6000,4800,0.5542893260294052
+6000,5100,0.697366690132294
+6000,5400,0.6305583913511819
+6000,5700,0.728727630242833
+6000,6000,1.0
diff --git a/metrics.jsonl b/metrics.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..9d06da4eb8e1fe04a963111a770f59986a276d6b
--- /dev/null
+++ b/metrics.jsonl
@@ -0,0 +1,661 @@
+{"step": 10, "train_loss": 2.766908884048462, "train_acc": 0.05944055944055944, "lr": 0.0006599999999999999}
+{"step": 20, "train_loss": 2.766866445541382, "train_acc": 0.05181347150259067, "lr": 0.00126}
+{"step": 30, "train_loss": 2.7672386169433594, "train_acc": 0.055077452667814115, "lr": 0.0018599999999999999}
+{"step": 40, "train_loss": 2.767467975616455, "train_acc": 0.06338028169014084, "lr": 0.00246}
+{"step": 50, "train_loss": 2.7671210765838623, "train_acc": 0.06382978723404255, "lr": 0.00306}
+{"step": 60, "train_loss": 2.767165422439575, "train_acc": 0.035211267605633804, "lr": 0.0036599999999999996}
+{"step": 70, "train_loss": 2.7671875953674316, "train_acc": 0.06166666666666667, "lr": 0.004259999999999999}
+{"step": 80, "train_loss": 2.7672226428985596, "train_acc": 0.07452339688041594, "lr": 0.00486}
+{"step": 90, "train_loss": 2.767040252685547, "train_acc": 0.06521739130434782, "lr": 0.00546}
+{"step": 100, "train_loss": 2.7664778232574463, "train_acc": 0.0721830985915493, "lr": 0.00606}
+{"step": 100, "val_loss": 2.7725763480545678, "val_acc": 0.06739316738079613}
+{"step": 110, "train_loss": 2.7670137882232666, "train_acc": 0.07222222222222222, "lr": 0.00666}
+{"step": 120, "train_loss": 2.7671194076538086, "train_acc": 0.06040268456375839, "lr": 0.007259999999999999}
+{"step": 130, "train_loss": 2.7668895721435547, "train_acc": 0.0683453237410072, "lr": 0.00786}
+{"step": 140, "train_loss": 2.7666685581207275, "train_acc": 0.07220216606498195, "lr": 0.008459999999999999}
+{"step": 150, "train_loss": 2.7667369842529297, "train_acc": 0.05291970802919708, "lr": 0.009059999999999999}
+{"step": 160, "train_loss": 2.767042875289917, "train_acc": 0.07441016333938294, "lr": 0.00966}
+{"step": 170, "train_loss": 2.7670507431030273, "train_acc": 0.055865921787709494, "lr": 0.01026}
+{"step": 180, "train_loss": 2.7668845653533936, "train_acc": 0.051457975986277875, "lr": 0.01086}
+{"step": 190, "train_loss": 2.766714572906494, "train_acc": 0.07653061224489796, "lr": 0.01146}
+{"step": 200, "train_loss": 2.76657772064209, "train_acc": 0.08362369337979095, "lr": 0.01206}
+{"step": 200, "val_loss": 2.7725299256066607, "val_acc": 0.06826654123185255}
+{"step": 210, "train_loss": 2.7670161724090576, "train_acc": 0.05136986301369863, "lr": 0.01266}
+{"step": 220, "train_loss": 2.767165422439575, "train_acc": 0.04225352112676056, "lr": 0.01326}
+{"step": 230, "train_loss": 2.7671170234680176, "train_acc": 0.06846846846846846, "lr": 0.01386}
+{"step": 240, "train_loss": 2.766571044921875, "train_acc": 0.05871559633027523, "lr": 0.014459999999999999}
+{"step": 250, "train_loss": 2.7669026851654053, "train_acc": 0.08006814310051108, "lr": 0.015059999999999999}
+{"step": 260, "train_loss": 2.7667016983032227, "train_acc": 0.061837455830388695, "lr": 0.01566}
+{"step": 270, "train_loss": 2.767120838165283, "train_acc": 0.08547008547008547, "lr": 0.01626}
+{"step": 280, "train_loss": 2.766993522644043, "train_acc": 0.07664233576642336, "lr": 0.01686}
+{"step": 290, "train_loss": 2.767467975616455, "train_acc": 0.0735042735042735, "lr": 0.01746}
+{"step": 300, "train_loss": 2.767251968383789, "train_acc": 0.05385996409335727, "lr": 0.01806}
+{"step": 300, "val_loss": 2.7724476687773003, "val_acc": 0.06804994428082085}
+{"step": 310, "train_loss": 2.76723313331604, "train_acc": 0.06174957118353345, "lr": 0.01866}
+{"step": 320, "train_loss": 2.766906976699829, "train_acc": 0.07008547008547009, "lr": 0.01926}
+{"step": 330, "train_loss": 2.7675414085388184, "train_acc": 0.06643952299829642, "lr": 0.01986}
+{"step": 340, "train_loss": 2.7674999237060547, "train_acc": 0.06260869565217392, "lr": 0.020460000000000002}
+{"step": 350, "train_loss": 2.76745343208313, "train_acc": 0.08156028368794327, "lr": 0.02106}
+{"step": 360, "train_loss": 2.7681946754455566, "train_acc": 0.07183364839319471, "lr": 0.02166}
+{"step": 370, "train_loss": 2.767892599105835, "train_acc": 0.061946902654867256, "lr": 0.02226}
+{"step": 380, "train_loss": 2.7682580947875977, "train_acc": 0.06851851851851852, "lr": 0.02286}
+{"step": 390, "train_loss": 2.768706798553467, "train_acc": 0.07439446366782007, "lr": 0.02346}
+{"step": 400, "train_loss": 2.7688870429992676, "train_acc": 0.05789473684210526, "lr": 0.02406}
+{"step": 400, "val_loss": 2.77230886984876, "val_acc": 0.06546853727674383}
+{"step": 410, "train_loss": 2.7688791751861572, "train_acc": 0.0757314974182444, "lr": 0.024659999999999998}
+{"step": 420, "train_loss": 2.7697010040283203, "train_acc": 0.06260869565217392, "lr": 0.025259999999999998}
+{"step": 430, "train_loss": 2.769599199295044, "train_acc": 0.06794425087108014, "lr": 0.025859999999999998}
+{"step": 440, "train_loss": 2.7696869373321533, "train_acc": 0.09108159392789374, "lr": 0.02646}
+{"step": 450, "train_loss": 2.769653797149658, "train_acc": 0.06097560975609756, "lr": 0.02706}
+{"step": 460, "train_loss": 2.7692790031433105, "train_acc": 0.05235602094240838, "lr": 0.02766}
+{"step": 470, "train_loss": 2.7693424224853516, "train_acc": 0.05018587360594796, "lr": 0.028259999999999997}
+{"step": 480, "train_loss": 2.768761157989502, "train_acc": 0.06093189964157706, "lr": 0.028859999999999997}
+{"step": 490, "train_loss": 2.7698264122009277, "train_acc": 0.0492091388400703, "lr": 0.02946}
+{"step": 500, "train_loss": 2.769723892211914, "train_acc": 0.05985915492957746, "lr": 0.03}
+{"step": 500, "val_loss": 2.772084470282358, "val_acc": 0.06475588491038901}
+{"step": 510, "train_loss": 2.769373893737793, "train_acc": 0.08389261744966443, "lr": 0.03}
+{"step": 520, "train_loss": 2.7698915004730225, "train_acc": 0.06914893617021277, "lr": 0.03}
+{"step": 530, "train_loss": 2.7702221870422363, "train_acc": 0.0663265306122449, "lr": 0.03}
+{"step": 540, "train_loss": 2.769411325454712, "train_acc": 0.06143344709897611, "lr": 0.03}
+{"step": 550, "train_loss": 2.770315170288086, "train_acc": 0.08623548922056384, "lr": 0.03}
+{"step": 560, "train_loss": 2.770132303237915, "train_acc": 0.08916083916083917, "lr": 0.03}
+{"step": 570, "train_loss": 2.770338773727417, "train_acc": 0.07076350093109869, "lr": 0.03}
+{"step": 580, "train_loss": 2.770328998565674, "train_acc": 0.06866197183098592, "lr": 0.03}
+{"step": 590, "train_loss": 2.770714044570923, "train_acc": 0.061619718309859156, "lr": 0.03}
+{"step": 600, "train_loss": 2.770204782485962, "train_acc": 0.09386973180076628, "lr": 0.03}
+{"step": 600, "val_loss": 2.771099938681928, "val_acc": 0.09020067215449655}
+{"step": 610, "train_loss": 2.7698543071746826, "train_acc": 0.09398496240601503, "lr": 0.03}
+{"step": 620, "train_loss": 2.768425941467285, "train_acc": 0.11423550087873462, "lr": 0.03}
+{"step": 630, "train_loss": 2.767578125, "train_acc": 0.11443661971830986, "lr": 0.03}
+{"step": 640, "train_loss": 2.772245168685913, "train_acc": 0.10951526032315978, "lr": 0.03}
+{"step": 650, "train_loss": 2.764951705932617, "train_acc": 0.1400359066427289, "lr": 0.03}
+{"step": 660, "train_loss": 2.764343738555908, "train_acc": 0.12299465240641712, "lr": 0.03}
+{"step": 670, "train_loss": 2.764280080795288, "train_acc": 0.13553113553113552, "lr": 0.03}
+{"step": 680, "train_loss": 2.7684059143066406, "train_acc": 0.09908256880733946, "lr": 0.03}
+{"step": 690, "train_loss": 2.7602341175079346, "train_acc": 0.12546125461254612, "lr": 0.03}
+{"step": 700, "train_loss": 2.760937452316284, "train_acc": 0.12173913043478261, "lr": 0.03}
+{"step": 700, "val_loss": 2.7611472348849633, "val_acc": 0.13328088036835475}
+{"step": 710, "train_loss": 2.7562713623046875, "train_acc": 0.14285714285714285, "lr": 0.03}
+{"step": 720, "train_loss": 2.762289047241211, "train_acc": 0.14417744916820702, "lr": 0.03}
+{"step": 730, "train_loss": 2.7532286643981934, "train_acc": 0.10714285714285714, "lr": 0.03}
+{"step": 740, "train_loss": 2.7389094829559326, "train_acc": 0.13111888111888112, "lr": 0.03}
+{"step": 750, "train_loss": 2.751345634460449, "train_acc": 0.1634446397188049, "lr": 0.03}
+{"step": 760, "train_loss": 2.7610573768615723, "train_acc": 0.1140819964349376, "lr": 0.03}
+{"step": 770, "train_loss": 2.7476813793182373, "train_acc": 0.1315345699831366, "lr": 0.03}
+{"step": 780, "train_loss": 2.7358598709106445, "train_acc": 0.13416815742397137, "lr": 0.03}
+{"step": 790, "train_loss": 2.7364425659179688, "train_acc": 0.14564831261101244, "lr": 0.03}
+{"step": 800, "train_loss": 2.734959840774536, "train_acc": 0.15508021390374332, "lr": 0.03}
+{"step": 800, "val_loss": 2.743599384417443, "val_acc": 0.1362242219445702}
+{"step": 810, "train_loss": 2.730637550354004, "train_acc": 0.14661654135338345, "lr": 0.03}
+{"step": 820, "train_loss": 2.72510027885437, "train_acc": 0.15151515151515152, "lr": 0.03}
+{"step": 830, "train_loss": 2.7394869327545166, "train_acc": 0.1407942238267148, "lr": 0.03}
+{"step": 840, "train_loss": 2.7385642528533936, "train_acc": 0.1539855072463768, "lr": 0.03}
+{"step": 850, "train_loss": 2.722966432571411, "train_acc": 0.1551094890510949, "lr": 0.03}
+{"step": 860, "train_loss": 2.7523176670074463, "train_acc": 0.15498154981549817, "lr": 0.03}
+{"step": 870, "train_loss": 2.7261061668395996, "train_acc": 0.12199312714776632, "lr": 0.03}
+{"step": 880, "train_loss": 2.695552110671997, "train_acc": 0.16728624535315986, "lr": 0.03}
+{"step": 890, "train_loss": 2.684502601623535, "train_acc": 0.15482695810564662, "lr": 0.03}
+{"step": 900, "train_loss": 2.716365337371826, "train_acc": 0.1421232876712329, "lr": 0.03}
+{"step": 900, "val_loss": 2.700008000193791, "val_acc": 0.14288004594467624}
+{"step": 910, "train_loss": 2.731410026550293, "train_acc": 0.10843373493975904, "lr": 0.03}
+{"step": 920, "train_loss": 2.697817087173462, "train_acc": 0.1276595744680851, "lr": 0.03}
+{"step": 930, "train_loss": 2.696057081222534, "train_acc": 0.10968921389396709, "lr": 0.03}
+{"step": 940, "train_loss": 2.7174057960510254, "train_acc": 0.13284132841328414, "lr": 0.03}
+{"step": 950, "train_loss": 2.6709587574005127, "train_acc": 0.14487632508833923, "lr": 0.03}
+{"step": 960, "train_loss": 2.6932573318481445, "train_acc": 0.11194029850746269, "lr": 0.03}
+{"step": 970, "train_loss": 2.66530704498291, "train_acc": 0.1504424778761062, "lr": 0.03}
+{"step": 980, "train_loss": 2.6580848693847656, "train_acc": 0.13109243697478992, "lr": 0.03}
+{"step": 990, "train_loss": 2.6331021785736084, "train_acc": 0.12433392539964476, "lr": 0.03}
+{"step": 1000, "train_loss": 2.6510868072509766, "train_acc": 0.1453287197231834, "lr": 0.03}
+{"step": 1000, "val_loss": 2.639930665437558, "val_acc": 0.14811166689320943}
+{"step": 1010, "train_loss": 2.6599583625793457, "train_acc": 0.14821763602251406, "lr": 0.03}
+{"step": 1020, "train_loss": 2.665640354156494, "train_acc": 0.11921708185053381, "lr": 0.03}
+{"step": 1030, "train_loss": 2.63663911819458, "train_acc": 0.15439856373429084, "lr": 0.03}
+{"step": 1040, "train_loss": 2.631218910217285, "train_acc": 0.14745762711864407, "lr": 0.03}
+{"step": 1050, "train_loss": 2.699805974960327, "train_acc": 0.10951526032315978, "lr": 0.03}
+{"step": 1060, "train_loss": 2.601384162902832, "train_acc": 0.1522491349480969, "lr": 0.03}
+{"step": 1070, "train_loss": 2.59802508354187, "train_acc": 0.14788732394366197, "lr": 0.03}
+{"step": 1080, "train_loss": 2.5445220470428467, "train_acc": 0.15905245346869712, "lr": 0.03}
+{"step": 1090, "train_loss": 2.583752393722534, "train_acc": 0.20109689213893966, "lr": 0.03}
+{"step": 1100, "train_loss": 2.546060800552368, "train_acc": 0.16087751371115175, "lr": 0.03}
+{"step": 1100, "val_loss": 2.5459947410084904, "val_acc": 0.16395359061589412}
+{"step": 1110, "train_loss": 2.5441131591796875, "train_acc": 0.16326530612244897, "lr": 0.03}
+{"step": 1120, "train_loss": 2.551809310913086, "train_acc": 0.16353383458646617, "lr": 0.03}
+{"step": 1130, "train_loss": 2.523454189300537, "train_acc": 0.18505942275042445, "lr": 0.03}
+{"step": 1140, "train_loss": 2.519562244415283, "train_acc": 0.15873015873015872, "lr": 0.03}
+{"step": 1150, "train_loss": 2.4974217414855957, "train_acc": 0.1717902350813743, "lr": 0.03}
+{"step": 1160, "train_loss": 2.4435577392578125, "train_acc": 0.19898819561551434, "lr": 0.03}
+{"step": 1170, "train_loss": 2.462434768676758, "train_acc": 0.20848708487084872, "lr": 0.03}
+{"step": 1180, "train_loss": 2.4487972259521484, "train_acc": 0.1474820143884892, "lr": 0.03}
+{"step": 1190, "train_loss": 2.41862154006958, "train_acc": 0.1875, "lr": 0.03}
+{"step": 1200, "train_loss": 2.3941028118133545, "train_acc": 0.20135363790186125, "lr": 0.03}
+{"step": 1200, "val_loss": 2.386614423153067, "val_acc": 0.18793860630838954}
+{"step": 1210, "train_loss": 2.4330053329467773, "train_acc": 0.17190388170055454, "lr": 0.03}
+{"step": 1220, "train_loss": 2.3821370601654053, "train_acc": 0.18101933216168717, "lr": 0.03}
+{"step": 1230, "train_loss": 2.398561954498291, "train_acc": 0.1785063752276867, "lr": 0.03}
+{"step": 1240, "train_loss": 2.3456857204437256, "train_acc": 0.2058319039451115, "lr": 0.03}
+{"step": 1250, "train_loss": 2.3747479915618896, "train_acc": 0.18927973199329984, "lr": 0.03}
+{"step": 1260, "train_loss": 2.3547372817993164, "train_acc": 0.17383512544802868, "lr": 0.03}
+{"step": 1270, "train_loss": 2.2840476036071777, "train_acc": 0.22522522522522523, "lr": 0.03}
+{"step": 1280, "train_loss": 2.284917116165161, "train_acc": 0.22336769759450173, "lr": 0.03}
+{"step": 1290, "train_loss": 2.2883901596069336, "train_acc": 0.19834710743801653, "lr": 0.03}
+{"step": 1300, "train_loss": 2.282930850982666, "train_acc": 0.20905923344947736, "lr": 0.03}
+{"step": 1300, "val_loss": 2.2622115347158185, "val_acc": 0.20377491004007517}
+{"step": 1310, "train_loss": 2.2334251403808594, "train_acc": 0.23037542662116042, "lr": 0.03}
+{"step": 1320, "train_loss": 2.2415146827697754, "train_acc": 0.2037037037037037, "lr": 0.03}
+{"step": 1330, "train_loss": 2.2246780395507812, "train_acc": 0.19897084048027444, "lr": 0.03}
+{"step": 1340, "train_loss": 2.235962152481079, "train_acc": 0.2010221465076661, "lr": 0.03}
+{"step": 1350, "train_loss": 2.2245540618896484, "train_acc": 0.216796875, "lr": 0.03}
+{"step": 1360, "train_loss": 2.2089242935180664, "train_acc": 0.23007246376811594, "lr": 0.03}
+{"step": 1370, "train_loss": 2.194188117980957, "train_acc": 0.2177121771217712, "lr": 0.03}
+{"step": 1380, "train_loss": 2.2158203125, "train_acc": 0.18904593639575973, "lr": 0.03}
+{"step": 1390, "train_loss": 2.150761842727661, "train_acc": 0.2226027397260274, "lr": 0.03}
+{"step": 1400, "train_loss": 2.1699352264404297, "train_acc": 0.2208904109589041, "lr": 0.03}
+{"step": 1400, "val_loss": 2.1675877007165942, "val_acc": 0.21690227284349953}
+{"step": 1410, "train_loss": 2.188222646713257, "train_acc": 0.23859649122807017, "lr": 0.03}
+{"step": 1420, "train_loss": 2.1307225227355957, "train_acc": 0.23877917414721722, "lr": 0.03}
+{"step": 1430, "train_loss": 2.1833181381225586, "train_acc": 0.2066420664206642, "lr": 0.03}
+{"step": 1440, "train_loss": 2.22863507270813, "train_acc": 0.2114695340501792, "lr": 0.03}
+{"step": 1450, "train_loss": 2.188725709915161, "train_acc": 0.23105360443622922, "lr": 0.03}
+{"step": 1460, "train_loss": 2.118905544281006, "train_acc": 0.2286212914485166, "lr": 0.03}
+{"step": 1470, "train_loss": 2.1393768787384033, "train_acc": 0.22021660649819494, "lr": 0.03}
+{"step": 1480, "train_loss": 2.1314048767089844, "train_acc": 0.20172413793103447, "lr": 0.03}
+{"step": 1490, "train_loss": 2.116321086883545, "train_acc": 0.2075134168157424, "lr": 0.03}
+{"step": 1500, "train_loss": 2.1887848377227783, "train_acc": 0.2014519056261343, "lr": 0.03}
+{"step": 1500, "val_loss": 2.100847718000978, "val_acc": 0.22441518005519928}
+{"step": 1510, "train_loss": 2.0444931983947754, "train_acc": 0.25249169435215946, "lr": 0.03}
+{"step": 1520, "train_loss": 2.0517306327819824, "train_acc": 0.2638888888888889, "lr": 0.03}
+{"step": 1530, "train_loss": 2.0862271785736084, "train_acc": 0.22580645161290322, "lr": 0.03}
+{"step": 1540, "train_loss": 2.1001482009887695, "train_acc": 0.22597864768683273, "lr": 0.03}
+{"step": 1550, "train_loss": 2.0849809646606445, "train_acc": 0.24914089347079038, "lr": 0.03}
+{"step": 1560, "train_loss": 1.9801485538482666, "train_acc": 0.2470389170896785, "lr": 0.03}
+{"step": 1570, "train_loss": 2.102132558822632, "train_acc": 0.22702702702702704, "lr": 0.03}
+{"step": 1580, "train_loss": 2.0454535484313965, "train_acc": 0.26173285198555957, "lr": 0.03}
+{"step": 1590, "train_loss": 2.018289804458618, "train_acc": 0.24006908462867013, "lr": 0.03}
+{"step": 1600, "train_loss": 2.03277587890625, "train_acc": 0.23774954627949182, "lr": 0.03}
+{"step": 1600, "val_loss": 2.0421807385743715, "val_acc": 0.23812685620566193}
+{"step": 1610, "train_loss": 2.068197250366211, "train_acc": 0.2427007299270073, "lr": 0.03}
+{"step": 1620, "train_loss": 2.058476686477661, "train_acc": 0.2222222222222222, "lr": 0.03}
+{"step": 1630, "train_loss": 2.0526340007781982, "train_acc": 0.23519163763066203, "lr": 0.03}
+{"step": 1640, "train_loss": 1.9723548889160156, "train_acc": 0.27704918032786885, "lr": 0.03}
+{"step": 1650, "train_loss": 2.140194892883301, "train_acc": 0.20947176684881602, "lr": 0.03}
+{"step": 1660, "train_loss": 2.0716552734375, "train_acc": 0.22142857142857142, "lr": 0.03}
+{"step": 1670, "train_loss": 2.1086809635162354, "train_acc": 0.18884892086330934, "lr": 0.03}
+{"step": 1680, "train_loss": 2.1454293727874756, "train_acc": 0.20147874306839186, "lr": 0.03}
+{"step": 1690, "train_loss": 2.0329196453094482, "train_acc": 0.2625, "lr": 0.03}
+{"step": 1700, "train_loss": 2.116617441177368, "train_acc": 0.2170940170940171, "lr": 0.03}
+{"step": 1700, "val_loss": 2.0225360929496548, "val_acc": 0.24167753085449986}
+{"step": 1710, "train_loss": 2.0198872089385986, "train_acc": 0.25668449197860965, "lr": 0.03}
+{"step": 1720, "train_loss": 2.1038858890533447, "train_acc": 0.21509433962264152, "lr": 0.03}
+{"step": 1730, "train_loss": 2.0211849212646484, "train_acc": 0.24956369982547993, "lr": 0.03}
+{"step": 1740, "train_loss": 2.11932373046875, "train_acc": 0.22831858407079647, "lr": 0.03}
+{"step": 1750, "train_loss": 2.072141647338867, "train_acc": 0.22727272727272727, "lr": 0.03}
+{"step": 1760, "train_loss": 2.1064934730529785, "train_acc": 0.2424812030075188, "lr": 0.03}
+{"step": 1770, "train_loss": 1.9349175691604614, "train_acc": 0.24740484429065743, "lr": 0.03}
+{"step": 1780, "train_loss": 2.04146146774292, "train_acc": 0.23722627737226276, "lr": 0.03}
+{"step": 1790, "train_loss": 1.917441725730896, "train_acc": 0.27164685908319186, "lr": 0.03}
+{"step": 1800, "train_loss": 2.0281143188476562, "train_acc": 0.26996197718631176, "lr": 0.03}
+{"step": 1800, "val_loss": 2.0046736210426555, "val_acc": 0.2532483899963113}
+{"step": 1810, "train_loss": 1.959876537322998, "train_acc": 0.2760416666666667, "lr": 0.03}
+{"step": 1820, "train_loss": 2.080301523208618, "train_acc": 0.22822299651567945, "lr": 0.03}
+{"step": 1830, "train_loss": 2.00892972946167, "train_acc": 0.24518388791593695, "lr": 0.03}
+{"step": 1840, "train_loss": 2.0326387882232666, "train_acc": 0.2427536231884058, "lr": 0.03}
+{"step": 1850, "train_loss": 1.9325389862060547, "train_acc": 0.30664395229982966, "lr": 0.03}
+{"step": 1860, "train_loss": 2.045886754989624, "train_acc": 0.241635687732342, "lr": 0.03}
+{"step": 1870, "train_loss": 2.0268497467041016, "train_acc": 0.23958333333333334, "lr": 0.03}
+{"step": 1880, "train_loss": 2.031115770339966, "train_acc": 0.2765957446808511, "lr": 0.03}
+{"step": 1890, "train_loss": 2.0654513835906982, "train_acc": 0.23443223443223443, "lr": 0.03}
+{"step": 1900, "train_loss": 2.060872793197632, "train_acc": 0.26258992805755393, "lr": 0.03}
+{"step": 1900, "val_loss": 1.9809743999875395, "val_acc": 0.25982231270884676}
+{"step": 1910, "train_loss": 2.0152580738067627, "train_acc": 0.24027072758037224, "lr": 0.03}
+{"step": 1920, "train_loss": 2.0540966987609863, "train_acc": 0.22784810126582278, "lr": 0.03}
+{"step": 1930, "train_loss": 1.892470121383667, "train_acc": 0.304421768707483, "lr": 0.03}
+{"step": 1940, "train_loss": 1.9371768236160278, "train_acc": 0.2864406779661017, "lr": 0.03}
+{"step": 1950, "train_loss": 2.033298969268799, "train_acc": 0.25217391304347825, "lr": 0.03}
+{"step": 1960, "train_loss": 1.883488655090332, "train_acc": 0.30404217926186294, "lr": 0.03}
+{"step": 1970, "train_loss": 1.8670023679733276, "train_acc": 0.32642487046632124, "lr": 0.03}
+{"step": 1980, "train_loss": 1.96165132522583, "train_acc": 0.27041742286751363, "lr": 0.03}
+{"step": 1990, "train_loss": 1.9298720359802246, "train_acc": 0.3016157989228007, "lr": 0.03}
+{"step": 2000, "train_loss": 1.906690239906311, "train_acc": 0.28862478777589134, "lr": 0.03}
+{"step": 2000, "val_loss": 1.9533350863201993, "val_acc": 0.2691142922543164}
+{"step": 2010, "train_loss": 2.0222008228302, "train_acc": 0.2595978062157221, "lr": 0.03}
+{"step": 2020, "train_loss": 1.9462976455688477, "train_acc": 0.2831858407079646, "lr": 0.03}
+{"step": 2030, "train_loss": 1.943488597869873, "train_acc": 0.2946593001841621, "lr": 0.03}
+{"step": 2040, "train_loss": 1.9127628803253174, "train_acc": 0.28620689655172415, "lr": 0.03}
+{"step": 2050, "train_loss": 1.927142858505249, "train_acc": 0.27708703374777977, "lr": 0.03}
+{"step": 2060, "train_loss": 1.9072448015213013, "train_acc": 0.3089887640449438, "lr": 0.03}
+{"step": 2070, "train_loss": 1.9241361618041992, "train_acc": 0.28091872791519434, "lr": 0.03}
+{"step": 2080, "train_loss": 1.9157605171203613, "train_acc": 0.28842832469775476, "lr": 0.03}
+{"step": 2090, "train_loss": 1.913435459136963, "train_acc": 0.287215411558669, "lr": 0.03}
+{"step": 2100, "train_loss": 1.9155278205871582, "train_acc": 0.28987993138936535, "lr": 0.03}
+{"step": 2100, "val_loss": 1.917813824530433, "val_acc": 0.2834801142090743}
+{"step": 2110, "train_loss": 1.9143223762512207, "train_acc": 0.2766355140186916, "lr": 0.03}
+{"step": 2120, "train_loss": 1.8066495656967163, "train_acc": 0.3121597096188748, "lr": 0.03}
+{"step": 2130, "train_loss": 1.892525553703308, "train_acc": 0.30726256983240224, "lr": 0.03}
+{"step": 2140, "train_loss": 1.8899070024490356, "train_acc": 0.3142857142857143, "lr": 0.03}
+{"step": 2150, "train_loss": 1.885166049003601, "train_acc": 0.29238754325259514, "lr": 0.03}
+{"step": 2160, "train_loss": 1.8982878923416138, "train_acc": 0.2712765957446808, "lr": 0.03}
+{"step": 2170, "train_loss": 2.0080995559692383, "train_acc": 0.2870201096892139, "lr": 0.03}
+{"step": 2180, "train_loss": 1.9485599994659424, "train_acc": 0.2929824561403509, "lr": 0.03}
+{"step": 2190, "train_loss": 1.986193299293518, "train_acc": 0.281767955801105, "lr": 0.03}
+{"step": 2200, "train_loss": 1.816292643547058, "train_acc": 0.33797909407665505, "lr": 0.03}
+{"step": 2200, "val_loss": 1.890189052260486, "val_acc": 0.2976994185731947}
+{"step": 2210, "train_loss": 1.923346996307373, "train_acc": 0.2905982905982906, "lr": 0.03}
+{"step": 2220, "train_loss": 1.898188829421997, "train_acc": 0.3058608058608059, "lr": 0.03}
+{"step": 2230, "train_loss": 1.8250718116760254, "train_acc": 0.31022530329289427, "lr": 0.03}
+{"step": 2240, "train_loss": 1.8951059579849243, "train_acc": 0.29182156133828996, "lr": 0.03}
+{"step": 2250, "train_loss": 1.9020919799804688, "train_acc": 0.3263888888888889, "lr": 0.03}
+{"step": 2260, "train_loss": 1.8743948936462402, "train_acc": 0.3207885304659498, "lr": 0.03}
+{"step": 2270, "train_loss": 1.8533838987350464, "train_acc": 0.3321678321678322, "lr": 0.03}
+{"step": 2280, "train_loss": 1.847572922706604, "train_acc": 0.3503787878787879, "lr": 0.03}
+{"step": 2290, "train_loss": 1.8434619903564453, "train_acc": 0.33157894736842103, "lr": 0.03}
+{"step": 2300, "train_loss": 1.836227297782898, "train_acc": 0.3516295025728988, "lr": 0.03}
+{"step": 2300, "val_loss": 1.8412683962583767, "val_acc": 0.33393595937420956}
+{"step": 2310, "train_loss": 1.828224778175354, "train_acc": 0.3546099290780142, "lr": 0.03}
+{"step": 2320, "train_loss": 1.843054175376892, "train_acc": 0.3486394557823129, "lr": 0.03}
+{"step": 2330, "train_loss": 1.7953587770462036, "train_acc": 0.36363636363636365, "lr": 0.03}
+{"step": 2340, "train_loss": 1.785981297492981, "train_acc": 0.3442028985507246, "lr": 0.03}
+{"step": 2350, "train_loss": 1.834628701210022, "train_acc": 0.32976827094474154, "lr": 0.03}
+{"step": 2360, "train_loss": 1.800986886024475, "train_acc": 0.343205574912892, "lr": 0.03}
+{"step": 2370, "train_loss": 1.8825178146362305, "train_acc": 0.3389830508474576, "lr": 0.03}
+{"step": 2380, "train_loss": 1.7201777696609497, "train_acc": 0.38564273789649417, "lr": 0.03}
+{"step": 2390, "train_loss": 1.8029751777648926, "train_acc": 0.3779385171790235, "lr": 0.03}
+{"step": 2400, "train_loss": 1.69386887550354, "train_acc": 0.4084507042253521, "lr": 0.03}
+{"step": 2400, "val_loss": 1.7344669714171865, "val_acc": 0.38473319342242895}
+{"step": 2410, "train_loss": 1.7288604974746704, "train_acc": 0.40874524714828897, "lr": 0.03}
+{"step": 2420, "train_loss": 1.7525873184204102, "train_acc": 0.39404553415061294, "lr": 0.03}
+{"step": 2430, "train_loss": 1.5940979719161987, "train_acc": 0.4315068493150685, "lr": 0.03}
+{"step": 2440, "train_loss": 1.592340350151062, "train_acc": 0.443859649122807, "lr": 0.03}
+{"step": 2450, "train_loss": 1.6096303462982178, "train_acc": 0.45036764705882354, "lr": 0.03}
+{"step": 2460, "train_loss": 1.6860132217407227, "train_acc": 0.45774647887323944, "lr": 0.03}
+{"step": 2470, "train_loss": 1.5813134908676147, "train_acc": 0.4484536082474227, "lr": 0.03}
+{"step": 2480, "train_loss": 1.6324824094772339, "train_acc": 0.44594594594594594, "lr": 0.03}
+{"step": 2490, "train_loss": 1.583082914352417, "train_acc": 0.4548736462093863, "lr": 0.03}
+{"step": 2500, "train_loss": 1.512461543083191, "train_acc": 0.4735915492957746, "lr": 0.03}
+{"step": 2500, "val_loss": 1.554759799376321, "val_acc": 0.48398210832139077}
+{"step": 2510, "train_loss": 1.4611777067184448, "train_acc": 0.505922165820643, "lr": 0.03}
+{"step": 2520, "train_loss": 1.456479549407959, "train_acc": 0.4972067039106145, "lr": 0.03}
+{"step": 2530, "train_loss": 1.5203490257263184, "train_acc": 0.5063752276867031, "lr": 0.03}
+{"step": 2540, "train_loss": 1.4402521848678589, "train_acc": 0.5083179297597042, "lr": 0.03}
+{"step": 2550, "train_loss": 1.4894459247589111, "train_acc": 0.5009140767824497, "lr": 0.03}
+{"step": 2560, "train_loss": 1.3851985931396484, "train_acc": 0.5672727272727273, "lr": 0.03}
+{"step": 2570, "train_loss": 1.4085540771484375, "train_acc": 0.5228070175438596, "lr": 0.03}
+{"step": 2580, "train_loss": 1.4504194259643555, "train_acc": 0.5247349823321554, "lr": 0.03}
+{"step": 2590, "train_loss": 1.3314634561538696, "train_acc": 0.5677179962894249, "lr": 0.03}
+{"step": 2600, "train_loss": 1.3229504823684692, "train_acc": 0.5737122557726465, "lr": 0.03}
+{"step": 2600, "val_loss": 1.3649557846328426, "val_acc": 0.5568576791619687}
+{"step": 2610, "train_loss": 1.3788292407989502, "train_acc": 0.5563258232235702, "lr": 0.03}
+{"step": 2620, "train_loss": 1.324334740638733, "train_acc": 0.5906642728904847, "lr": 0.03}
+{"step": 2630, "train_loss": 1.414539098739624, "train_acc": 0.560530679933665, "lr": 0.03}
+{"step": 2640, "train_loss": 1.3651878833770752, "train_acc": 0.5435606060606061, "lr": 0.03}
+{"step": 2650, "train_loss": 1.16623854637146, "train_acc": 0.6333907056798623, "lr": 0.03}
+{"step": 2660, "train_loss": 1.2599848508834839, "train_acc": 0.5921501706484642, "lr": 0.03}
+{"step": 2670, "train_loss": 1.1537723541259766, "train_acc": 0.6261343012704175, "lr": 0.03}
+{"step": 2680, "train_loss": 1.2112598419189453, "train_acc": 0.5960854092526691, "lr": 0.03}
+{"step": 2690, "train_loss": 1.1304254531860352, "train_acc": 0.6360544217687075, "lr": 0.03}
+{"step": 2700, "train_loss": 1.1994757652282715, "train_acc": 0.6020761245674741, "lr": 0.03}
+{"step": 2700, "val_loss": 1.1882815293720363, "val_acc": 0.6204762888724401}
+{"step": 2710, "train_loss": 1.145301103591919, "train_acc": 0.6476014760147601, "lr": 0.03}
+{"step": 2720, "train_loss": 1.188015341758728, "train_acc": 0.6260720411663808, "lr": 0.03}
+{"step": 2730, "train_loss": 1.0838757753372192, "train_acc": 0.6660714285714285, "lr": 0.03}
+{"step": 2740, "train_loss": 1.1726112365722656, "train_acc": 0.6385767790262172, "lr": 0.03}
+{"step": 2750, "train_loss": 1.2164965867996216, "train_acc": 0.6226053639846744, "lr": 0.03}
+{"step": 2760, "train_loss": 1.1693085432052612, "train_acc": 0.6178571428571429, "lr": 0.03}
+{"step": 2770, "train_loss": 1.0484473705291748, "train_acc": 0.6504504504504505, "lr": 0.03}
+{"step": 2780, "train_loss": 0.9667451977729797, "train_acc": 0.7033333333333334, "lr": 0.03}
+{"step": 2790, "train_loss": 1.1443794965744019, "train_acc": 0.6455026455026455, "lr": 0.03}
+{"step": 2800, "train_loss": 1.091179609298706, "train_acc": 0.6614310645724258, "lr": 0.03}
+{"step": 2800, "val_loss": 1.0647491137864504, "val_acc": 0.6624931748730427}
+{"step": 2810, "train_loss": 1.0932612419128418, "train_acc": 0.6605504587155964, "lr": 0.03}
+{"step": 2820, "train_loss": 0.9682689905166626, "train_acc": 0.6824561403508772, "lr": 0.03}
+{"step": 2830, "train_loss": 1.0750550031661987, "train_acc": 0.6838365896980462, "lr": 0.03}
+{"step": 2840, "train_loss": 0.9423609375953674, "train_acc": 0.6833333333333333, "lr": 0.03}
+{"step": 2850, "train_loss": 0.8859216570854187, "train_acc": 0.7212543554006968, "lr": 0.03}
+{"step": 2860, "train_loss": 0.9749433994293213, "train_acc": 0.6959459459459459, "lr": 0.03}
+{"step": 2870, "train_loss": 0.8630385398864746, "train_acc": 0.7423971377459749, "lr": 0.03}
+{"step": 2880, "train_loss": 0.9621990919113159, "train_acc": 0.70578231292517, "lr": 0.03}
+{"step": 2890, "train_loss": 0.9465541243553162, "train_acc": 0.7076124567474048, "lr": 0.03}
+{"step": 2900, "train_loss": 0.9025654196739197, "train_acc": 0.7235494880546075, "lr": 0.03}
+{"step": 2900, "val_loss": 0.9671723400893588, "val_acc": 0.6938616355282773}
+{"step": 2910, "train_loss": 0.8868045210838318, "train_acc": 0.7107583774250441, "lr": 0.03}
+{"step": 2920, "train_loss": 0.9943187236785889, "train_acc": 0.6910112359550562, "lr": 0.03}
+{"step": 2930, "train_loss": 0.9027404189109802, "train_acc": 0.7223198594024605, "lr": 0.03}
+{"step": 2940, "train_loss": 0.8541110157966614, "train_acc": 0.7234782608695652, "lr": 0.03}
+{"step": 2950, "train_loss": 0.9322529435157776, "train_acc": 0.7171532846715328, "lr": 0.03}
+{"step": 2960, "train_loss": 0.8315759897232056, "train_acc": 0.7375886524822695, "lr": 0.03}
+{"step": 2970, "train_loss": 0.7862561345100403, "train_acc": 0.7530864197530864, "lr": 0.03}
+{"step": 2980, "train_loss": 0.771465003490448, "train_acc": 0.75, "lr": 0.03}
+{"step": 2990, "train_loss": 0.746562659740448, "train_acc": 0.7700729927007299, "lr": 0.03}
+{"step": 3000, "train_loss": 0.7951160669326782, "train_acc": 0.760539629005059, "lr": 0.03}
+{"step": 3000, "val_loss": 0.8140730856855232, "val_acc": 0.7394545783492398}
+{"step": 3010, "train_loss": 0.7464771270751953, "train_acc": 0.756993006993007, "lr": 0.03}
+{"step": 3020, "train_loss": 0.7385881543159485, "train_acc": 0.7532956685499058, "lr": 0.03}
+{"step": 3030, "train_loss": 0.6035742163658142, "train_acc": 0.8057921635434412, "lr": 0.03}
+{"step": 3040, "train_loss": 0.7275730967521667, "train_acc": 0.7791304347826087, "lr": 0.03}
+{"step": 3050, "train_loss": 0.6897069215774536, "train_acc": 0.7724867724867724, "lr": 0.03}
+{"step": 3060, "train_loss": 0.7843987345695496, "train_acc": 0.7670753064798599, "lr": 0.03}
+{"step": 3070, "train_loss": 0.7025309205055237, "train_acc": 0.7771739130434783, "lr": 0.03}
+{"step": 3080, "train_loss": 0.7334039211273193, "train_acc": 0.7624309392265194, "lr": 0.03}
+{"step": 3090, "train_loss": 0.740171492099762, "train_acc": 0.7609108159392789, "lr": 0.03}
+{"step": 3100, "train_loss": 0.7316293716430664, "train_acc": 0.7808695652173913, "lr": 0.03}
+{"step": 3100, "val_loss": 0.7016682670509969, "val_acc": 0.781723383827289}
+{"step": 3110, "train_loss": 0.6091770529747009, "train_acc": 0.8050089445438283, "lr": 0.03}
+{"step": 3120, "train_loss": 0.605404257774353, "train_acc": 0.7986348122866894, "lr": 0.03}
+{"step": 3130, "train_loss": 0.5877877473831177, "train_acc": 0.8131313131313131, "lr": 0.03}
+{"step": 3140, "train_loss": 0.5835033655166626, "train_acc": 0.82, "lr": 0.03}
+{"step": 3150, "train_loss": 0.5859155058860779, "train_acc": 0.8017241379310345, "lr": 0.03}
+{"step": 3160, "train_loss": 0.6010102033615112, "train_acc": 0.7954545454545454, "lr": 0.03}
+{"step": 3170, "train_loss": 0.614943265914917, "train_acc": 0.7777777777777778, "lr": 0.03}
+{"step": 3180, "train_loss": 0.6161959171295166, "train_acc": 0.8205574912891986, "lr": 0.03}
+{"step": 3190, "train_loss": 0.5507194399833679, "train_acc": 0.8167539267015707, "lr": 0.03}
+{"step": 3200, "train_loss": 0.5717699527740479, "train_acc": 0.8178633975481612, "lr": 0.03}
+{"step": 3200, "val_loss": 0.5645294462323347, "val_acc": 0.8256069738778978}
+{"step": 3210, "train_loss": 0.567050576210022, "train_acc": 0.8057921635434412, "lr": 0.03}
+{"step": 3220, "train_loss": 0.5601882338523865, "train_acc": 0.8169491525423729, "lr": 0.03}
+{"step": 3230, "train_loss": 0.44248083233833313, "train_acc": 0.8498293515358362, "lr": 0.03}
+{"step": 3240, "train_loss": 0.46875256299972534, "train_acc": 0.8501742160278746, "lr": 0.03}
+{"step": 3250, "train_loss": 0.5752190947532654, "train_acc": 0.8341793570219966, "lr": 0.03}
+{"step": 3260, "train_loss": 0.5420324206352234, "train_acc": 0.8405017921146953, "lr": 0.03}
+{"step": 3270, "train_loss": 0.46700140833854675, "train_acc": 0.8512544802867383, "lr": 0.03}
+{"step": 3280, "train_loss": 0.5198332071304321, "train_acc": 0.8392523364485981, "lr": 0.03}
+{"step": 3290, "train_loss": 0.4908716380596161, "train_acc": 0.8597785977859779, "lr": 0.03}
+{"step": 3300, "train_loss": 0.47613686323165894, "train_acc": 0.8513986013986014, "lr": 0.03}
+{"step": 3300, "val_loss": 0.4791334604055987, "val_acc": 0.8466097892953184}
+{"step": 3310, "train_loss": 0.5462414026260376, "train_acc": 0.8286713286713286, "lr": 0.03}
+{"step": 3320, "train_loss": 0.45297959446907043, "train_acc": 0.8739205526770294, "lr": 0.03}
+{"step": 3330, "train_loss": 0.42478814721107483, "train_acc": 0.8499156829679595, "lr": 0.03}
+{"step": 3340, "train_loss": 0.47758373618125916, "train_acc": 0.8541666666666666, "lr": 0.03}
+{"step": 3350, "train_loss": 0.38148877024650574, "train_acc": 0.8628472222222222, "lr": 0.03}
+{"step": 3360, "train_loss": 0.3638796806335449, "train_acc": 0.892226148409894, "lr": 0.03}
+{"step": 3370, "train_loss": 0.4068819284439087, "train_acc": 0.8592592592592593, "lr": 0.03}
+{"step": 3380, "train_loss": 0.4821833372116089, "train_acc": 0.8594306049822064, "lr": 0.03}
+{"step": 3390, "train_loss": 0.43418368697166443, "train_acc": 0.8620689655172413, "lr": 0.03}
+{"step": 3400, "train_loss": 0.3031265437602997, "train_acc": 0.9045936395759717, "lr": 0.03}
+{"step": 3400, "val_loss": 0.3796190192069994, "val_acc": 0.877382621061786}
+{"step": 3410, "train_loss": 0.37388697266578674, "train_acc": 0.8860103626943006, "lr": 0.03}
+{"step": 3420, "train_loss": 0.37773481011390686, "train_acc": 0.8765432098765432, "lr": 0.03}
+{"step": 3430, "train_loss": 0.3322020173072815, "train_acc": 0.8959435626102292, "lr": 0.03}
+{"step": 3440, "train_loss": 0.3630320727825165, "train_acc": 0.8908045977011494, "lr": 0.03}
+{"step": 3450, "train_loss": 0.3548167943954468, "train_acc": 0.8854545454545455, "lr": 0.03}
+{"step": 3460, "train_loss": 0.3259003460407257, "train_acc": 0.9012567324955116, "lr": 0.03}
+{"step": 3470, "train_loss": 0.4095349907875061, "train_acc": 0.8821490467937608, "lr": 0.03}
+{"step": 3480, "train_loss": 0.3697845935821533, "train_acc": 0.8756756756756757, "lr": 0.03}
+{"step": 3490, "train_loss": 0.3445562720298767, "train_acc": 0.8986254295532646, "lr": 0.03}
+{"step": 3500, "train_loss": 0.4108409583568573, "train_acc": 0.8407534246575342, "lr": 0.03}
+{"step": 3500, "val_loss": 0.3573944267445617, "val_acc": 0.8909240502363629}
+{"step": 3510, "train_loss": 0.3336193561553955, "train_acc": 0.8986254295532646, "lr": 0.03}
+{"step": 3520, "train_loss": 0.3458668887615204, "train_acc": 0.8930581613508443, "lr": 0.03}
+{"step": 3530, "train_loss": 0.5662059783935547, "train_acc": 0.8872987477638641, "lr": 0.03}
+{"step": 3540, "train_loss": 0.27768903970718384, "train_acc": 0.9090909090909091, "lr": 0.03}
+{"step": 3550, "train_loss": 0.28785574436187744, "train_acc": 0.9130434782608695, "lr": 0.03}
+{"step": 3560, "train_loss": 0.3107292354106903, "train_acc": 0.9097222222222222, "lr": 0.03}
+{"step": 3570, "train_loss": 0.28930702805519104, "train_acc": 0.9003496503496503, "lr": 0.03}
+{"step": 3580, "train_loss": 0.33099955320358276, "train_acc": 0.8971119133574007, "lr": 0.03}
+{"step": 3590, "train_loss": 0.248138427734375, "train_acc": 0.9432013769363167, "lr": 0.03}
+{"step": 3600, "train_loss": 0.25592100620269775, "train_acc": 0.9203980099502488, "lr": 0.03}
+{"step": 3600, "val_loss": 0.27603237870701997, "val_acc": 0.9135014538066221}
+{"step": 3610, "train_loss": 0.3000580072402954, "train_acc": 0.910873440285205, "lr": 0.03}
+{"step": 3620, "train_loss": 0.26131513714790344, "train_acc": 0.9219858156028369, "lr": 0.03}
+{"step": 3630, "train_loss": 0.30908170342445374, "train_acc": 0.9139579349904398, "lr": 0.03}
+{"step": 3640, "train_loss": 0.2777124345302582, "train_acc": 0.9102564102564102, "lr": 0.03}
+{"step": 3650, "train_loss": 0.25701266527175903, "train_acc": 0.9319852941176471, "lr": 0.03}
+{"step": 3660, "train_loss": 0.28932711482048035, "train_acc": 0.9128919860627178, "lr": 0.03}
+{"step": 3670, "train_loss": 0.34869763255119324, "train_acc": 0.9111111111111111, "lr": 0.03}
+{"step": 3680, "train_loss": 0.2934812605381012, "train_acc": 0.9208261617900172, "lr": 0.03}
+{"step": 3690, "train_loss": 0.22039029002189636, "train_acc": 0.9265232974910395, "lr": 0.03}
+{"step": 3700, "train_loss": 0.3259594738483429, "train_acc": 0.9071428571428571, "lr": 0.03}
+{"step": 3700, "val_loss": 0.24830701795828197, "val_acc": 0.9223177923751752}
+{"step": 3710, "train_loss": 0.2720879018306732, "train_acc": 0.9108910891089109, "lr": 0.03}
+{"step": 3720, "train_loss": 0.2697032690048218, "train_acc": 0.9243542435424354, "lr": 0.03}
+{"step": 3730, "train_loss": 0.23536936938762665, "train_acc": 0.9321739130434783, "lr": 0.03}
+{"step": 3740, "train_loss": 0.2792978882789612, "train_acc": 0.9190647482014388, "lr": 0.03}
+{"step": 3750, "train_loss": 0.2738732695579529, "train_acc": 0.9212454212454212, "lr": 0.03}
+{"step": 3760, "train_loss": 0.23929283022880554, "train_acc": 0.9232142857142858, "lr": 0.03}
+{"step": 3770, "train_loss": 0.2697831392288208, "train_acc": 0.9323308270676691, "lr": 0.03}
+{"step": 3780, "train_loss": 0.2586655914783478, "train_acc": 0.9405594405594405, "lr": 0.03}
+{"step": 3790, "train_loss": 0.2045179009437561, "train_acc": 0.9452991452991453, "lr": 0.03}
+{"step": 3800, "train_loss": 0.2414364516735077, "train_acc": 0.9213286713286714, "lr": 0.03}
+{"step": 3800, "val_loss": 0.23910732279559888, "val_acc": 0.9277975565595912}
+{"step": 3810, "train_loss": 0.229457825422287, "train_acc": 0.9275862068965517, "lr": 0.03}
+{"step": 3820, "train_loss": 0.16436830163002014, "train_acc": 0.9521367521367521, "lr": 0.03}
+{"step": 3830, "train_loss": 0.2693682312965393, "train_acc": 0.916955017301038, "lr": 0.03}
+{"step": 3840, "train_loss": 0.2048877328634262, "train_acc": 0.9412844036697248, "lr": 0.03}
+{"step": 3850, "train_loss": 0.283812940120697, "train_acc": 0.9214145383104125, "lr": 0.03}
+{"step": 3860, "train_loss": 0.17520830035209656, "train_acc": 0.9490333919156415, "lr": 0.03}
+{"step": 3870, "train_loss": 0.25150343775749207, "train_acc": 0.931095406360424, "lr": 0.03}
+{"step": 3880, "train_loss": 0.21367943286895752, "train_acc": 0.9395017793594306, "lr": 0.03}
+{"step": 3890, "train_loss": 0.2304462492465973, "train_acc": 0.9282945736434108, "lr": 0.03}
+{"step": 3900, "train_loss": 0.19886243343353271, "train_acc": 0.9334500875656743, "lr": 0.03}
+{"step": 3900, "val_loss": 0.2039344919603799, "val_acc": 0.9358333961274982}
+{"step": 3910, "train_loss": 0.29478713870048523, "train_acc": 0.9323943661971831, "lr": 0.03}
+{"step": 3920, "train_loss": 0.2147415578365326, "train_acc": 0.9305301645338209, "lr": 0.03}
+{"step": 3930, "train_loss": 0.15721093118190765, "train_acc": 0.955595026642984, "lr": 0.03}
+{"step": 3940, "train_loss": 0.2628743648529053, "train_acc": 0.925, "lr": 0.03}
+{"step": 3950, "train_loss": 0.14344939589500427, "train_acc": 0.959409594095941, "lr": 0.03}
+{"step": 3960, "train_loss": 0.1777331382036209, "train_acc": 0.9429065743944637, "lr": 0.03}
+{"step": 3970, "train_loss": 0.18083584308624268, "train_acc": 0.9402985074626866, "lr": 0.03}
+{"step": 3980, "train_loss": 0.19037257134914398, "train_acc": 0.9362831858407079, "lr": 0.03}
+{"step": 3990, "train_loss": 0.1969563215970993, "train_acc": 0.922945205479452, "lr": 0.03}
+{"step": 4000, "train_loss": 0.22336672246456146, "train_acc": 0.9346289752650176, "lr": 0.03}
+{"step": 4000, "val_loss": 0.19565565219572373, "val_acc": 0.9411204819481068}
+{"step": 4010, "train_loss": 0.21513959765434265, "train_acc": 0.9320214669051878, "lr": 0.03}
+{"step": 4020, "train_loss": 0.17559082806110382, "train_acc": 0.9506172839506173, "lr": 0.03}
+{"step": 4030, "train_loss": 0.11849553883075714, "train_acc": 0.9676840215439856, "lr": 0.03}
+{"step": 4040, "train_loss": 0.21440282464027405, "train_acc": 0.9470802919708029, "lr": 0.03}
+{"step": 4050, "train_loss": 0.2309122085571289, "train_acc": 0.9460431654676259, "lr": 0.03}
+{"step": 4060, "train_loss": 0.1715807020664215, "train_acc": 0.948905109489051, "lr": 0.03}
+{"step": 4070, "train_loss": 0.17208929359912872, "train_acc": 0.9436860068259386, "lr": 0.03}
+{"step": 4080, "train_loss": 0.1687222123146057, "train_acc": 0.9425925925925925, "lr": 0.03}
+{"step": 4090, "train_loss": 0.27178579568862915, "train_acc": 0.9221014492753623, "lr": 0.03}
+{"step": 4100, "train_loss": 0.11636045575141907, "train_acc": 0.9636048526863085, "lr": 0.03}
+{"step": 4100, "val_loss": 0.1830405669002447, "val_acc": 0.9478914079799975}
+{"step": 4110, "train_loss": 0.23218518495559692, "train_acc": 0.9364791288566243, "lr": 0.03}
+{"step": 4120, "train_loss": 0.21447838842868805, "train_acc": 0.9500924214417745, "lr": 0.03}
+{"step": 4130, "train_loss": 0.1393199861049652, "train_acc": 0.9537671232876712, "lr": 0.03}
+{"step": 4140, "train_loss": 0.18842875957489014, "train_acc": 0.9496644295302014, "lr": 0.03}
+{"step": 4150, "train_loss": 0.16264431178569794, "train_acc": 0.9515260323159784, "lr": 0.03}
+{"step": 4160, "train_loss": 0.17930492758750916, "train_acc": 0.9492119089316988, "lr": 0.03}
+{"step": 4170, "train_loss": 0.1726948767900467, "train_acc": 0.9531531531531532, "lr": 0.03}
+{"step": 4180, "train_loss": 0.16865624487400055, "train_acc": 0.9459962756052142, "lr": 0.03}
+{"step": 4190, "train_loss": 0.13902616500854492, "train_acc": 0.949740034662045, "lr": 0.03}
+{"step": 4200, "train_loss": 0.16583965718746185, "train_acc": 0.9484902309058615, "lr": 0.03}
+{"step": 4200, "val_loss": 0.163355083023342, "val_acc": 0.9524514201788853}
+{"step": 4210, "train_loss": 0.23131629824638367, "train_acc": 0.935251798561151, "lr": 0.03}
+{"step": 4220, "train_loss": 0.2106121927499771, "train_acc": 0.9370503597122302, "lr": 0.03}
+{"step": 4230, "train_loss": 0.14810703694820404, "train_acc": 0.9602169981916817, "lr": 0.03}
+{"step": 4240, "train_loss": 0.12294060736894608, "train_acc": 0.958029197080292, "lr": 0.03}
+{"step": 4250, "train_loss": 0.13806426525115967, "train_acc": 0.9591836734693877, "lr": 0.03}
+{"step": 4260, "train_loss": 0.1987415999174118, "train_acc": 0.9523809523809523, "lr": 0.03}
+{"step": 4270, "train_loss": 0.17750221490859985, "train_acc": 0.9402460456942003, "lr": 0.03}
+{"step": 4280, "train_loss": 0.19475196301937103, "train_acc": 0.9385964912280702, "lr": 0.03}
+{"step": 4290, "train_loss": 0.11116267740726471, "train_acc": 0.9617391304347827, "lr": 0.03}
+{"step": 4300, "train_loss": 0.12821601331233978, "train_acc": 0.9501779359430605, "lr": 0.03}
+{"step": 4300, "val_loss": 0.13511455139979558, "val_acc": 0.9634762366452589}
+{"step": 4310, "train_loss": 0.08713290095329285, "train_acc": 0.9767441860465116, "lr": 0.03}
+{"step": 4320, "train_loss": 0.13724355399608612, "train_acc": 0.9534482758620689, "lr": 0.03}
+{"step": 4330, "train_loss": 0.1567355990409851, "train_acc": 0.9502487562189055, "lr": 0.03}
+{"step": 4340, "train_loss": 0.11165188252925873, "train_acc": 0.9590747330960854, "lr": 0.03}
+{"step": 4350, "train_loss": 0.15741369128227234, "train_acc": 0.946917808219178, "lr": 0.03}
+{"step": 4360, "train_loss": 0.18265314400196075, "train_acc": 0.947935368043088, "lr": 0.03}
+{"step": 4370, "train_loss": 0.14089681208133698, "train_acc": 0.9546247818499127, "lr": 0.03}
+{"step": 4380, "train_loss": 0.08998037874698639, "train_acc": 0.9744058500914077, "lr": 0.03}
+{"step": 4390, "train_loss": 0.15314096212387085, "train_acc": 0.9568965517241379, "lr": 0.03}
+{"step": 4400, "train_loss": 0.16435973346233368, "train_acc": 0.9468864468864469, "lr": 0.03}
+{"step": 4400, "val_loss": 0.1412610718447364, "val_acc": 0.960945353365234}
+{"step": 4410, "train_loss": 0.13116709887981415, "train_acc": 0.9543057996485061, "lr": 0.03}
+{"step": 4420, "train_loss": 0.20207273960113525, "train_acc": 0.9421052631578948, "lr": 0.03}
+{"step": 4430, "train_loss": 0.1186913251876831, "train_acc": 0.9617391304347827, "lr": 0.03}
+{"step": 4440, "train_loss": 0.16798478364944458, "train_acc": 0.956140350877193, "lr": 0.03}
+{"step": 4450, "train_loss": 0.12218335270881653, "train_acc": 0.9618573797678275, "lr": 0.03}
+{"step": 4460, "train_loss": 0.08962336927652359, "train_acc": 0.9678638941398866, "lr": 0.03}
+{"step": 4470, "train_loss": 0.1374485045671463, "train_acc": 0.9476439790575916, "lr": 0.03}
+{"step": 4480, "train_loss": 0.12414844334125519, "train_acc": 0.9602169981916817, "lr": 0.03}
+{"step": 4490, "train_loss": 0.14917001128196716, "train_acc": 0.9507042253521126, "lr": 0.03}
+{"step": 4500, "train_loss": 0.09270349144935608, "train_acc": 0.9738219895287958, "lr": 0.03}
+{"step": 4500, "val_loss": 0.1191985422446712, "val_acc": 0.9646285331257151}
+{"step": 4510, "train_loss": 0.14716050028800964, "train_acc": 0.9535315985130112, "lr": 0.03}
+{"step": 4520, "train_loss": 0.13523012399673462, "train_acc": 0.9597902097902098, "lr": 0.03}
+{"step": 4530, "train_loss": 0.11574624478816986, "train_acc": 0.9595782073813708, "lr": 0.03}
+{"step": 4540, "train_loss": 0.1791483461856842, "train_acc": 0.9492753623188406, "lr": 0.03}
+{"step": 4550, "train_loss": 0.08854511380195618, "train_acc": 0.9767025089605734, "lr": 0.03}
+{"step": 4560, "train_loss": 0.11759722977876663, "train_acc": 0.9552238805970149, "lr": 0.03}
+{"step": 4570, "train_loss": 0.116185262799263, "train_acc": 0.9646182495344506, "lr": 0.03}
+{"step": 4580, "train_loss": 0.12946707010269165, "train_acc": 0.9467353951890034, "lr": 0.03}
+{"step": 4590, "train_loss": 0.1327163726091385, "train_acc": 0.9517470881863561, "lr": 0.03}
+{"step": 4600, "train_loss": 0.10477261245250702, "train_acc": 0.9731663685152058, "lr": 0.03}
+{"step": 4600, "val_loss": 0.11069907310623699, "val_acc": 0.9694421317935777}
+{"step": 4610, "train_loss": 0.11202438175678253, "train_acc": 0.9655797101449275, "lr": 0.03}
+{"step": 4620, "train_loss": 0.08453045785427094, "train_acc": 0.9775086505190311, "lr": 0.03}
+{"step": 4630, "train_loss": 0.0891466736793518, "train_acc": 0.975177304964539, "lr": 0.03}
+{"step": 4640, "train_loss": 0.11938533186912537, "train_acc": 0.9629629629629629, "lr": 0.03}
+{"step": 4650, "train_loss": 0.1094445213675499, "train_acc": 0.9668411867364747, "lr": 0.03}
+{"step": 4660, "train_loss": 0.10334756970405579, "train_acc": 0.972972972972973, "lr": 0.03}
+{"step": 4670, "train_loss": 0.11996053904294968, "train_acc": 0.9664804469273743, "lr": 0.03}
+{"step": 4680, "train_loss": 0.08701901882886887, "train_acc": 0.9728353140916808, "lr": 0.03}
+{"step": 4690, "train_loss": 0.14446045458316803, "train_acc": 0.9648798521256932, "lr": 0.03}
+{"step": 4700, "train_loss": 0.0843559205532074, "train_acc": 0.9750415973377704, "lr": 0.03}
+{"step": 4700, "val_loss": 0.09771193633809648, "val_acc": 0.9708814772334164}
+{"step": 4710, "train_loss": 0.19518014788627625, "train_acc": 0.9572368421052632, "lr": 0.03}
+{"step": 4720, "train_loss": 0.11867877095937729, "train_acc": 0.9692028985507246, "lr": 0.03}
+{"step": 4730, "train_loss": 0.11774397641420364, "train_acc": 0.9659090909090909, "lr": 0.03}
+{"step": 4740, "train_loss": 0.11106429249048233, "train_acc": 0.9683098591549296, "lr": 0.03}
+{"step": 4750, "train_loss": 0.1005198135972023, "train_acc": 0.9715242881072027, "lr": 0.03}
+{"step": 4760, "train_loss": 0.10715421289205551, "train_acc": 0.9717314487632509, "lr": 0.03}
+{"step": 4770, "train_loss": 0.13624192774295807, "train_acc": 0.9617391304347827, "lr": 0.03}
+{"step": 4780, "train_loss": 0.11759638786315918, "train_acc": 0.9635416666666666, "lr": 0.03}
+{"step": 4790, "train_loss": 0.0845421552658081, "train_acc": 0.9738675958188153, "lr": 0.03}
+{"step": 4800, "train_loss": 0.12653589248657227, "train_acc": 0.9598603839441536, "lr": 0.03}
+{"step": 4800, "val_loss": 0.09127685628562666, "val_acc": 0.9703219397723474}
+{"step": 4810, "train_loss": 0.09926635771989822, "train_acc": 0.9682804674457429, "lr": 0.03}
+{"step": 4820, "train_loss": 0.07466734945774078, "train_acc": 0.9753787878787878, "lr": 0.03}
+{"step": 4830, "train_loss": 0.10839094966650009, "train_acc": 0.96875, "lr": 0.03}
+{"step": 4840, "train_loss": 0.08932317793369293, "train_acc": 0.9759704251386322, "lr": 0.03}
+{"step": 4850, "train_loss": 0.15319226682186127, "train_acc": 0.9536082474226805, "lr": 0.03}
+{"step": 4860, "train_loss": 0.1283373087644577, "train_acc": 0.9621621621621622, "lr": 0.03}
+{"step": 4870, "train_loss": 0.08430033177137375, "train_acc": 0.9743119266055046, "lr": 0.03}
+{"step": 4880, "train_loss": 0.13086159527301788, "train_acc": 0.9616788321167883, "lr": 0.03}
+{"step": 4890, "train_loss": 0.14240580797195435, "train_acc": 0.9586206896551724, "lr": 0.03}
+{"step": 4900, "train_loss": 0.09506326913833618, "train_acc": 0.9725557461406518, "lr": 0.03}
+{"step": 4900, "val_loss": 0.1168869048310824, "val_acc": 0.9649437211282461}
+{"step": 4910, "train_loss": 0.1497807651758194, "train_acc": 0.95, "lr": 0.03}
+{"step": 4920, "train_loss": 0.07703424245119095, "train_acc": 0.979557069846678, "lr": 0.03}
+{"step": 4930, "train_loss": 0.09255093336105347, "train_acc": 0.9738317757009346, "lr": 0.03}
+{"step": 4940, "train_loss": 0.09539364278316498, "train_acc": 0.9680284191829485, "lr": 0.03}
+{"step": 4950, "train_loss": 0.11073766648769379, "train_acc": 0.9597806215722121, "lr": 0.03}
+{"step": 4960, "train_loss": 0.1399698555469513, "train_acc": 0.9574829931972789, "lr": 0.03}
+{"step": 4970, "train_loss": 0.09922947734594345, "train_acc": 0.9753954305799648, "lr": 0.03}
+{"step": 4980, "train_loss": 0.07084730267524719, "train_acc": 0.9738317757009346, "lr": 0.03}
+{"step": 4990, "train_loss": 0.11004573851823807, "train_acc": 0.9705882352941176, "lr": 0.03}
+{"step": 5000, "train_loss": 0.11262394487857819, "train_acc": 0.9597701149425287, "lr": 0.03}
+{"step": 5000, "val_loss": 0.09772448670635794, "val_acc": 0.971313827983302}
+{"step": 5010, "train_loss": 0.08323325216770172, "train_acc": 0.9735915492957746, "lr": 0.03}
+{"step": 5020, "train_loss": 0.09223984181880951, "train_acc": 0.9759704251386322, "lr": 0.03}
+{"step": 5030, "train_loss": 0.11447209864854813, "train_acc": 0.9651567944250871, "lr": 0.03}
+{"step": 5040, "train_loss": 0.17013421654701233, "train_acc": 0.9493243243243243, "lr": 0.03}
+{"step": 5050, "train_loss": 0.09760969132184982, "train_acc": 0.9743589743589743, "lr": 0.03}
+{"step": 5060, "train_loss": 0.09790698438882828, "train_acc": 0.974910394265233, "lr": 0.03}
+{"step": 5070, "train_loss": 0.08855703473091125, "train_acc": 0.9816666666666667, "lr": 0.03}
+{"step": 5080, "train_loss": 0.11070235073566437, "train_acc": 0.9638989169675091, "lr": 0.03}
+{"step": 5090, "train_loss": 0.054000794887542725, "train_acc": 0.9827882960413081, "lr": 0.03}
+{"step": 5100, "train_loss": 0.09287554770708084, "train_acc": 0.966789667896679, "lr": 0.03}
+{"step": 5100, "val_loss": 0.10964289947398377, "val_acc": 0.9668808474525932}
+{"step": 5110, "train_loss": 0.06902170181274414, "train_acc": 0.9805309734513274, "lr": 0.03}
+{"step": 5120, "train_loss": 0.12709848582744598, "train_acc": 0.9569120287253142, "lr": 0.03}
+{"step": 5130, "train_loss": 0.07549445331096649, "train_acc": 0.9805996472663139, "lr": 0.03}
+{"step": 5140, "train_loss": 0.11690355837345123, "train_acc": 0.9678510998307953, "lr": 0.03}
+{"step": 5150, "train_loss": 0.12338673323392868, "train_acc": 0.9621928166351607, "lr": 0.03}
+{"step": 5160, "train_loss": 0.08981168270111084, "train_acc": 0.9717138103161398, "lr": 0.03}
+{"step": 5170, "train_loss": 0.10502207279205322, "train_acc": 0.9727427597955707, "lr": 0.03}
+{"step": 5180, "train_loss": 0.07035193592309952, "train_acc": 0.9781420765027322, "lr": 0.03}
+{"step": 5190, "train_loss": 0.11336053162813187, "train_acc": 0.9611992945326279, "lr": 0.03}
+{"step": 5200, "train_loss": 0.11169255524873734, "train_acc": 0.9701754385964912, "lr": 0.03}
+{"step": 5200, "val_loss": 0.07819674487324374, "val_acc": 0.9780353762877667}
+{"step": 5210, "train_loss": 0.09326975047588348, "train_acc": 0.9802867383512545, "lr": 0.03}
+{"step": 5220, "train_loss": 0.0857771784067154, "train_acc": 0.98, "lr": 0.03}
+{"step": 5230, "train_loss": 0.09606002271175385, "train_acc": 0.9646840148698885, "lr": 0.03}
+{"step": 5240, "train_loss": 0.1070253774523735, "train_acc": 0.9704861111111112, "lr": 0.03}
+{"step": 5250, "train_loss": 0.09058088809251785, "train_acc": 0.9720767888307156, "lr": 0.03}
+{"step": 5260, "train_loss": 0.0995791107416153, "train_acc": 0.9734982332155477, "lr": 0.03}
+{"step": 5270, "train_loss": 0.08250360190868378, "train_acc": 0.9736842105263158, "lr": 0.03}
+{"step": 5280, "train_loss": 0.10103976726531982, "train_acc": 0.96875, "lr": 0.03}
+{"step": 5290, "train_loss": 0.09552822262048721, "train_acc": 0.9709401709401709, "lr": 0.03}
+{"step": 5300, "train_loss": 0.07611660659313202, "train_acc": 0.9743589743589743, "lr": 0.03}
+{"step": 5300, "val_loss": 0.07911717389537982, "val_acc": 0.9763406052056472}
+{"step": 5310, "train_loss": 0.07065031677484512, "train_acc": 0.9842931937172775, "lr": 0.03}
+{"step": 5320, "train_loss": 0.05751626193523407, "train_acc": 0.9781420765027322, "lr": 0.03}
+{"step": 5330, "train_loss": 0.13463662564754486, "train_acc": 0.9674502712477396, "lr": 0.03}
+{"step": 5340, "train_loss": 0.07787031680345535, "train_acc": 0.9773519163763066, "lr": 0.03}
+{"step": 5350, "train_loss": 0.082365982234478, "train_acc": 0.9786856127886323, "lr": 0.03}
+{"step": 5360, "train_loss": 0.097504623234272, "train_acc": 0.9725776965265083, "lr": 0.03}
+{"step": 5370, "train_loss": 0.0790698304772377, "train_acc": 0.9773123909249564, "lr": 0.03}
+{"step": 5380, "train_loss": 0.08751873672008514, "train_acc": 0.9719789842381786, "lr": 0.03}
+{"step": 5390, "train_loss": 0.08116444945335388, "train_acc": 0.9715370018975332, "lr": 0.03}
+{"step": 5400, "train_loss": 0.14442434906959534, "train_acc": 0.963963963963964, "lr": 0.03}
+{"step": 5400, "val_loss": 0.10916551518487001, "val_acc": 0.967337485088352}
+{"step": 5410, "train_loss": 0.11909222602844238, "train_acc": 0.9585585585585585, "lr": 0.03}
+{"step": 5420, "train_loss": 0.10424216836690903, "train_acc": 0.9690346083788707, "lr": 0.03}
+{"step": 5430, "train_loss": 0.07699041813611984, "train_acc": 0.9753954305799648, "lr": 0.03}
+{"step": 5440, "train_loss": 0.10567527264356613, "train_acc": 0.9616087751371115, "lr": 0.03}
+{"step": 5450, "train_loss": 0.04888017848134041, "train_acc": 0.989247311827957, "lr": 0.03}
+{"step": 5460, "train_loss": 0.08176710456609726, "train_acc": 0.9768270944741533, "lr": 0.03}
+{"step": 5470, "train_loss": 0.07789792865514755, "train_acc": 0.9755244755244755, "lr": 0.03}
+{"step": 5480, "train_loss": 0.1290452778339386, "train_acc": 0.9647058823529412, "lr": 0.03}
+{"step": 5490, "train_loss": 0.14932133257389069, "train_acc": 0.9516129032258065, "lr": 0.03}
+{"step": 5500, "train_loss": 0.08361385762691498, "train_acc": 0.9726775956284153, "lr": 0.03}
+{"step": 5500, "val_loss": 0.07804288066536204, "val_acc": 0.9751213313206817}
+{"step": 5510, "train_loss": 0.10949866473674774, "train_acc": 0.9593023255813954, "lr": 0.03}
+{"step": 5520, "train_loss": 0.11483132839202881, "train_acc": 0.9660107334525939, "lr": 0.03}
+{"step": 5530, "train_loss": 0.061465054750442505, "train_acc": 0.9826086956521739, "lr": 0.03}
+{"step": 5540, "train_loss": 0.06778427213430405, "train_acc": 0.9712230215827338, "lr": 0.03}
+{"step": 5550, "train_loss": 0.06739257276058197, "train_acc": 0.9782971619365609, "lr": 0.03}
+{"step": 5560, "train_loss": 0.050486136227846146, "train_acc": 0.9805996472663139, "lr": 0.03}
+{"step": 5570, "train_loss": 0.1036604642868042, "train_acc": 0.9664310954063604, "lr": 0.03}
+{"step": 5580, "train_loss": 0.09106454998254776, "train_acc": 0.9675213675213675, "lr": 0.03}
+{"step": 5590, "train_loss": 0.10779085010290146, "train_acc": 0.9744990892531876, "lr": 0.03}
+{"step": 5600, "train_loss": 0.053984660655260086, "train_acc": 0.9857651245551602, "lr": 0.03}
+{"step": 5600, "val_loss": 0.10134145603274344, "val_acc": 0.9755667917777531}
+{"step": 5610, "train_loss": 0.06807176768779755, "train_acc": 0.9771528998242531, "lr": 0.03}
+{"step": 5620, "train_loss": 0.0796903595328331, "train_acc": 0.9754816112084063, "lr": 0.03}
+{"step": 5630, "train_loss": 0.103957898914814, "train_acc": 0.9688644688644689, "lr": 0.03}
+{"step": 5640, "train_loss": 0.11236396431922913, "train_acc": 0.9637681159420289, "lr": 0.03}
+{"step": 5650, "train_loss": 0.08244230598211288, "train_acc": 0.9722735674676525, "lr": 0.03}
+{"step": 5660, "train_loss": 0.07498233765363693, "train_acc": 0.9765765765765766, "lr": 0.03}
+{"step": 5670, "train_loss": 0.07900435477495193, "train_acc": 0.9724137931034482, "lr": 0.03}
+{"step": 5680, "train_loss": 0.059358786791563034, "train_acc": 0.9875, "lr": 0.03}
+{"step": 5690, "train_loss": 0.08152535557746887, "train_acc": 0.9767857142857143, "lr": 0.03}
+{"step": 5700, "train_loss": 0.08548393100500107, "train_acc": 0.9695431472081218, "lr": 0.03}
+{"step": 5700, "val_loss": 0.0728262062025563, "val_acc": 0.9766726077257492}
+{"step": 5710, "train_loss": 0.07796569168567657, "train_acc": 0.9676258992805755, "lr": 0.03}
+{"step": 5720, "train_loss": 0.09751331061124802, "train_acc": 0.9703832752613241, "lr": 0.03}
+{"step": 5730, "train_loss": 0.1745457947254181, "train_acc": 0.9629629629629629, "lr": 0.03}
+{"step": 5740, "train_loss": 0.08052671700716019, "train_acc": 0.9718804920913884, "lr": 0.03}
+{"step": 5750, "train_loss": 0.0847795307636261, "train_acc": 0.9809358752166378, "lr": 0.03}
+{"step": 5760, "train_loss": 0.10679459571838379, "train_acc": 0.9661921708185054, "lr": 0.03}
+{"step": 5770, "train_loss": 0.06960411369800568, "train_acc": 0.9786096256684492, "lr": 0.03}
+{"step": 5780, "train_loss": 0.09565853327512741, "train_acc": 0.9676840215439856, "lr": 0.03}
+{"step": 5790, "train_loss": 0.04611029475927353, "train_acc": 0.9851576994434137, "lr": 0.03}
+{"step": 5800, "train_loss": 0.06185757368803024, "train_acc": 0.9857904085257548, "lr": 0.03}
+{"step": 5800, "val_loss": 0.07573951228117516, "val_acc": 0.9772400285783863}
+{"step": 5810, "train_loss": 0.07216297090053558, "train_acc": 0.9755244755244755, "lr": 0.03}
+{"step": 5820, "train_loss": 0.08344211429357529, "train_acc": 0.9720767888307156, "lr": 0.03}
+{"step": 5830, "train_loss": 0.05572304129600525, "train_acc": 0.9874326750448833, "lr": 0.03}
+{"step": 5840, "train_loss": 0.05596315115690231, "train_acc": 0.9775280898876404, "lr": 0.03}
+{"step": 5850, "train_loss": 0.1067180410027504, "train_acc": 0.9681978798586572, "lr": 0.03}
+{"step": 5860, "train_loss": 0.07980313897132874, "train_acc": 0.9769503546099291, "lr": 0.03}
+{"step": 5870, "train_loss": 0.07813811302185059, "train_acc": 0.9782214156079855, "lr": 0.03}
+{"step": 5880, "train_loss": 0.06460016220808029, "train_acc": 0.9776632302405498, "lr": 0.03}
+{"step": 5890, "train_loss": 0.09999273717403412, "train_acc": 0.962756052141527, "lr": 0.03}
+{"step": 5900, "train_loss": 0.06985883414745331, "train_acc": 0.9860383944153578, "lr": 0.03}
+{"step": 5900, "val_loss": 0.07268840765841339, "val_acc": 0.9785573912048786}
+{"step": 5910, "train_loss": 0.056510090827941895, "train_acc": 0.9881956155143339, "lr": 0.03}
+{"step": 5920, "train_loss": 0.07709793001413345, "train_acc": 0.9773123909249564, "lr": 0.03}
+{"step": 5930, "train_loss": 0.05471811816096306, "train_acc": 0.9860869565217392, "lr": 0.03}
+{"step": 5940, "train_loss": 0.05578899011015892, "train_acc": 0.9806338028169014, "lr": 0.03}
+{"step": 5950, "train_loss": 0.0880856066942215, "train_acc": 0.975925925925926, "lr": 0.03}
+{"step": 5960, "train_loss": 0.14970485866069794, "train_acc": 0.957089552238806, "lr": 0.03}
+{"step": 5970, "train_loss": 0.062378063797950745, "train_acc": 0.9796610169491525, "lr": 0.03}
+{"step": 5980, "train_loss": 0.0321226604282856, "train_acc": 0.993006993006993, "lr": 0.03}
+{"step": 5990, "train_loss": 0.11230643838644028, "train_acc": 0.9696428571428571, "lr": 0.03}
+{"step": 6000, "train_loss": 0.06138728931546211, "train_acc": 0.9836363636363636, "lr": 0.03}
+{"step": 6000, "val_loss": 0.07559233404324017, "val_acc": 0.9789880997470876}
+{"step": 6000, "final_val_loss": 0.07559233404324017, "final_val_acc": 0.9789880997470876}
diff --git a/path_decomp_trajectory/path_pair_matrices/global_tn_sims_from_paths.npz b/path_decomp_trajectory/path_pair_matrices/global_tn_sims_from_paths.npz
new file mode 100644
index 0000000000000000000000000000000000000000..5f8e82481f06ab212be5dc3de35c017575f29c27
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/global_tn_sims_from_paths.npz
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32903b7532f5e1d6dd60936c24288a3ac9ad21236a515fe1e1a73d021c40aa22
+size 9525
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_all_layer_blocks.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_all_layer_blocks.png
new file mode 100644
index 0000000000000000000000000000000000000000..3c0b4a27acc5b753fde2adc5301f0f12c2441db9
Binary files /dev/null and b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_all_layer_blocks.png differ
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_direct.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_direct.png
new file mode 100644
index 0000000000000000000000000000000000000000..cb63a88dd801270e8f786ddff07ccc50ea15b44f
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_direct.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f471d5a1c3d8b72687dca56cc5c3547137f39dd32ba7e0c84fcf81bb5757735
+size 115508
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_layer1.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_layer1.png
new file mode 100644
index 0000000000000000000000000000000000000000..3bab5fa82fa7724e87b35893fd328d2b598f26e6
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_layer1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2772856e89dab9ac5cfacebf37b6109667b6e6e76ffd04969e7ff437ffb2db8
+size 115276
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_layer2.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_layer2.png
new file mode 100644
index 0000000000000000000000000000000000000000..44eeea6323810eac63111eef9e9b751781f4a50f
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_direct_vs_layer2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bafcf80fd99a2b1ed32eb077a6baebb5d598f2f7cf5d055a149665cfebfa606b
+size 115709
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_direct.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_direct.png
new file mode 100644
index 0000000000000000000000000000000000000000..c4785e2955cc50772ad17074787712362bef809f
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_direct.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45927eca893e3d4d465afdb1b95c5e28d5cb2b63673d56e5b429e8f7f887e5ab
+size 115293
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_layer1.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_layer1.png
new file mode 100644
index 0000000000000000000000000000000000000000..9851c42709e3968670666e91d6d0231f8e7ba2ba
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_layer1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:317e2237cab9d57114c4faedea74890c09afbd3271e3226b11d5b2733e3f150f
+size 115622
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_layer2.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_layer2.png
new file mode 100644
index 0000000000000000000000000000000000000000..63d19af9fc58bd2bfab4dd666380d0657f8aa387
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer1_vs_layer2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3ad436c497054c562b7a164476faaad759b8e0fd9039aa6dac155356b80b28c
+size 115694
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_direct.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_direct.png
new file mode 100644
index 0000000000000000000000000000000000000000..d7cfa9ce3ea8c0e752846353614539836dfc3370
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_direct.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c68c0b1adb2cef21012e135aad741f49f6560c612414e46c818e2ac116fb648
+size 115473
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_layer1.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_layer1.png
new file mode 100644
index 0000000000000000000000000000000000000000..7445273eff2cecd23eb468840e7c62bfc79b74a9
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_layer1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4eef7d5b28471d068cb0228b6dbaf3aede6f387dc5ca267da64dbd6f7e14a8c1
+size 115690
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_layer2.png b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_layer2.png
new file mode 100644
index 0000000000000000000000000000000000000000..953f256499471dfb93a8fd1dc1369cd406c9329b
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_layer2_vs_layer2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0042f1a485245e6e1b16861f39072ba2e0d4df5d0fc265bc687257d097c0dcea
+size 116065
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_summary.csv b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_summary.csv
new file mode 100644
index 0000000000000000000000000000000000000000..8f87a7fc84db002a6c8dcc29383d4cb6ddb6fd51
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_heatmaps/local_summary.csv
@@ -0,0 +1,10 @@
+source_group,target_group,mean,min,max,finite_count
+direct,direct,0.8710301107112558,0.19711512620295,1.0,441
+direct,layer1,-0.1899674453251921,-0.3100421709556807,0.05320398251335634,441
+direct,layer2,-0.006856381402757138,-0.11606242462666022,0.08417227603030221,441
+layer1,direct,-0.18996744532519208,-0.3100421709556807,0.05320398251335634,441
+layer1,layer1,0.6019092668715977,-0.09097721103571482,1.0,441
+layer1,layer2,0.07253879821336423,-0.16236400702997092,0.28493605308693903,441
+layer2,direct,-0.006856381402757137,-0.11606242462666022,0.08417227603030221,441
+layer2,layer1,0.07253879821336423,-0.16236400702997092,0.28493605308693903,441
+layer2,layer2,0.21889750434179336,-0.16715588251437632,1.0,441
diff --git a/path_decomp_trajectory/path_pair_matrices/layer_block_tn_sims.npz b/path_decomp_trajectory/path_pair_matrices/layer_block_tn_sims.npz
new file mode 100644
index 0000000000000000000000000000000000000000..8b4bdab701e4dd7fb97d445334986764b9223931
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/layer_block_tn_sims.npz
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc42051761585dec4ea8279e4d39f2fb1f38f2e7fbfce1f156ce221b08984388
+size 67875
diff --git a/path_decomp_trajectory/path_pair_matrices/path_pair_tn_matrices.npz b/path_decomp_trajectory/path_pair_matrices/path_pair_tn_matrices.npz
new file mode 100644
index 0000000000000000000000000000000000000000..db9c68b7c678e9fadafbcdbe14d01439d52a4ed8
--- /dev/null
+++ b/path_decomp_trajectory/path_pair_matrices/path_pair_tn_matrices.npz
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5374a943b0d6da216a71d7b10d0ccff8c769aaac897050e8d161fc09e99849a
+size 8159344