full TTT

pr_number	parameters
30	—
64	{"epochs":25,"learning_rate":0.012,"momentum":0.9,"freeze_blocks":0}
150	{"learning_rate":0.002,"epochs":3,"freeze_first_blocks":2}
152	{"learning_rate":0.002,"epochs":2,"momentum":0.9,"batch_size":32}
178	{"epochs":2,"learning_rate":0.002,"frozen_blocks":4}
223	{"learning_rate":0.0003,"epochs":1,"momentum":0.95}
254	{"learning_rate":0.002,"momentum":0.9,"epochs":3,"freezing_first_blocks":2}
263	{"learning_rate":0.0001,"steps":1,"scope":["attn.proj.weight","mlp.proj.weight"]}
264	{"optimizer":"SGD","learning_rate":0.002,"momentum":0.9,"epochs":2}
281	{"learning_rate":0.002,"momentum":0.9,"epochs":3}
290	{"epochs":3,"learning_rate":0.002,"freeze_blocks":2}
297	{"learning_rate":0.0003,"momentum":0.95}
301	{"epochs":3,"learning_rate":0.002,"momentum":0.9,"freeze_blocks":2}
302	{"scope":"MLP weights in last 3 blocks","learning_rate":null,"decay_prior":true}
303	{"learning_rate":0.002,"epochs":3,"freeze_blocks":2,"momentum":0.9,"gradient_clipping":1}
317	{"learning_rate":0.002,"epochs":3,"momentum":0.9,"freeze_blocks":2}
334	{"epochs":3,"frozen_blocks":2}
338	{"epochs":3,"learning_rate":0.002,"momentum":0.9,"frozen_blocks":2}
364	{"learning_rate":0.005,"momentum":0.9,"epochs":15,"freeze_blocks":0,"batch_seqs":16}
366	{"epochs":3,"learning_rate":0.002,"momentum":0.9,"grad_clip":1,"frozen_blocks":2}
384	{"learning_rate":0.002,"momentum":0.9,"epochs":2}
388	{"learning_rate":0.008,"epochs":25,"momentum":0.9,"batch_seqs":32,"freeze_blocks":0}
390	{"epochs":8,"learning_rate":0.002,"momentum":0.9}
397	{"learning_rate":0.002,"epochs":3,"freeze_blocks":2,"momentum":0.9}
398	{"epochs":20,"learning_rate":0.008,"momentum":0.9,"freeze_blocks":0}
401	{"epochs":8,"learning_rate":0.002,"momentum":0.9}
421	{"epochs":3,"optimizer":"SGD","time":"83s"}
442	{"learning_rate":0.0005,"epochs":10,"optimizer":"AdamW"}
455	{"epochs":3,"optimizer":"SGD","momentum":0.9,"learning_rate":0.002,"batch_size":32,"freezes_first_blocks":2}
481	{"optimizer":"AdamW","learning_rate":0.0005,"epochs":30,"cosine_decay":true,"per_layer_lr":true,"freeze_blocks":0,"batch_seqs_per_gpu":64}
486	{"optimizer":"AdamW","epochs":10,"freeze_blocks":0,"time_seconds":154}
509	{"epochs":30,"lr_schedule":"cosine decay"}
517	{"epochs":100,"learning_rate":0.001,"lr_min":0.00001,"scheduler":"cosine annealing"}
518	{"epochs":50,"learning_rate":0.0005,"weight_decay":0,"all_parameters_unfrozen":true,"per_layer_lr":{"mlp.proj":3,"mlp.fc":0.5},"grad_clip":1,"ddp_gradient_sync":true}
562	{"epochs":22,"optimizer":"AdamW","learning_rate":0.0005,"weight_decay":0,"lr_schedule":"per-step cosine decay to 0","per_layer_lr_groups":{"output_projections":3,"input_projections":0.5},"batch_size_per_gpu":32,"gradient_sync":"all_reduce per step","gradient_clipping":1,"TTT_time_seconds":406,"eval_time_seconds":197}
581	{"epochs":20,"learning_rate":0.0005,"min_learning_rate":0.00002}
587	—
595	{"optimizer":"AdamW","learning_rate":0.0005,"epochs":10,"weight_decay":0,"gradient_clipping":1}
639	{"optimizers_tested":["AdamW","SGD"],"learning_rates":[0.0005,0.002,0.001],"epochs":[3,5,10],"effect":"neutral-to-harmful on GPTQ weights"}
661	{"epochs":30,"schedule":"cosine","seed":1337}
667	{"learning_rate":0.003,"momentum":0.95,"epochs":3,"chunk_tokens":32768}
672	{"epochs":30,"optimizer":"AdamW","learning_rate":0.0005,"lr_schedule":"cosine decay","per_layer_lr_groups":{"mlp.proj":3,"mlp.fc":0.5}}
684	{"epochs":20}
685	{"phases":2,"phase_1":"cosine recovery","phase_2":"multi-pass score-first scoring","passes":3}
686	{"learning_rate":0.002,"epochs":3,"chunk_tokens":32768,"freeze_blocks":2,"untie":false}
696	{"learning_rate":0.002,"epochs_per_chunk":2,"stride":256,"chunk_tokens":32768,"batch_seqs":32,"all_parameters_adapt":true}
702	{"epochs":100,"learning_rate":0.001}
741	{"epochs":20,"learning_rate_schedule":"cosine","per_layer_lr_groups":true}
756	{"learning_rate":0.002,"epochs":3,"chunk_tokens":32768,"stride":64}
771	{"learning_rate":0.0005,"epochs":30,"cosine":true,"per_layer_lr":true,"freeze_blocks":0,"batch_seqs":64,"max_steps":300}
785	{"learning_rate":0.002,"epochs":3,"chunk_tokens":32768,"freeze_blocks":0,"momentum":0.9,"batch_seqs":32,"grad_clip":1}
852	{"epochs":3}
857	{"epochs":20,"learning_rate":0.0005}
928	{"enabled":1}
964	{"document_isolated":true,"reset_at_bos":true}
965	{"learning_rate":0.002,"epochs":3,"chunk_tokens":32768,"freeze_blocks":0,"momentum":0.9,"batch_seqs":32,"grad_clip":1}
967	{"epochs":4,"zero_frozen_blocks":true,"skip_sliding_eval":true}
977	{"legal":true}
1002	{"learning_rate":0.002,"epochs":3,"legal_score_first":true}
1005	{"learning_rate":0.002,"epochs":3,"chunk_tokens":32768}
1006	{"mode":"pre-quantization","epochs":3}
1009	{"learning_rate":0.002,"epochs":3}
1043	{"epochs":3,"learning_rate":0.002,"momentum":0.9,"grad_clip":1}
1045	{"learning_rate":0.002,"epochs":3}
1050	{"enabled":true}
1052	{"chunk":65536}
1077	{"window_size":32768,"optimizer":"SGD"}
1107	{"epochs":1}
1108	{"learning_rate_range":[0.00005,0.002]}
1186	{"epochs":20,"learning_rate":null}
1191	—
1193	—
1216	{"learning_rate":0.002,"epochs":3,"legal":true,"score_first":true}
1228	—
1255	{"optimizer":"AdamW"}
1257	{"learning_rate":0.0005,"epochs":3}
1270	{"epochs":3}
1275	—
1306	{"epochs":6,"learning_rate":0.0005,"freeze_first_blocks":2,"batch_size":32}
1310	{"learning_rate":0.002,"epochs":3,"all_blocks_unfrozen":true}
1318	{"learning_rate":0.001,"epochs":1,"frozen_blocks":"0-9"}
1328	{"optimizer":"AdamW","epochs":1,"learning_rate":0.001,"freeze_blocks":10}
1329	{"optimizer":"AdamW","epochs":1,"learning_rate":0.001,"freeze_blocks":10}
1351	{"timing":"pre-quantization","epochs":10,"freeze":0,"per_block_lr_scaling":{"start":0.3,"end":1,"interpolation":"linear"}}
1364	{"epochs":6,"freeze_first_blocks":2,"learning_rate_start":0.0005,"learning_rate_end":0.00005}
1376	{"epochs":6,"freeze_first_blocks":2,"learning_rate":0.0005}
1397	—
1400	{"epochs":10,"adaptive_lr":true,"per_block_lr":true}
1407	{"learning_rate":0.002,"epochs":3,"chunk_tokens":32768,"freeze_blocks":0,"momentum":0.9,"batch_seqs":32,"grad_clip":1}
1408	{"learning_rate":0.0005,"epochs":10,"freeze_blocks":0,"per_block_lr_scale":"0.3x-1.0x"}
1416	{"optimizer":"AdamW","epochs":6,"timing":"pre-quant"}
1422	{"optimizer":"SGD","learning_rate":0.005,"momentum":0.9,"chunk_size":2048,"all_weights":true}
1423	{"pre_quant":true,"epochs":6,"learning_rate":0.0005,"freeze_first_blocks":2}
1424	{"learning_rate":0.002,"epochs":3,"chunk_tokens":32768,"freeze_blocks":0,"momentum":0.9,"batch_seqs":32,"grad_clip":1}
1430	{"optimizer":"AdamW","epochs":1,"learning_rate":0.001,"freeze_blocks":"0-9","second_pass_fraction":0.1,"floor_lr":0.0001}
1482	{"epochs":8,"learning_rate":0.00045,"freeze_blocks":1}
1487	{"epochs":10,"learning_rate":0.00045,"freeze_blocks":1,"schedule":"cosine"}
1489	{"learning_rate":0.0005,"epochs":6,"freeze_blocks":2,"batch_seqs":32,"grad_clip":1}
1517	{"epochs":18,"learning_rate":0.0003,"freeze_blocks":1}
1529	{"enabled":true,"learning_rate":0.01}
1539	{"epochs":6,"learning_rate":0.0005,"freeze_blocks":2,"compiled":true}
1550	{"learning_rate":0.0005,"epochs":6,"freeze_blocks":2,"batch_size":32,"sequence_length":2048,"compiled":true}
1601	{"every":4,"inner_outer_split":"cross-chunk","delta_loss":true,"optimizer":"SAM","rho":0.05}
1620	{"micro_batching":true}
1624	{"learning_rate":0.00045,"epochs":10,"freeze_blocks":1}
1625	{"mode":"E2E","scope":"MLP-only in last fraction of blocks","last_frac":null,"learning_rate":0.015,"epochs":2}
1633	{"learning_rate":0.00045,"epochs":10,"freeze_blocks":1}
1638	{"adaptive_epochs":true,"max_epochs":null,"min_epochs":null,"ema":null}
1676	{"score_before_update":true,"single_pass":true,"learning_rate":0.005,"momentum":0.9,"epochs_per_chunk":3}
1697	{"learning_rate":0.005,"epochs":3}
1701	{"learning_rate":0.005,"epochs":3}
1702	{"easy_chunk_ratio":0.998,"easy_chunk_epochs":1,"outlier_drop_fraction":0.03,"score_weight_power":0.5}
1703	{"learning_rate":0.005,"epochs":3}
1735	{"enabled":true,"epochs":21,"parallel_gpus":8,"pre_quant":true}
1738	{"epochs":21,"parallel_gpus":8,"federated_averaging":true}
1758	{"learning_rate":0.001,"freeze_blocks":0,"epochs":21,"phase":"pre-quant"}
1760	{"epochs":7,"learning_rate":0.0005,"pre_quantization":true}