Open-weight model
sihc-research-checkpoints
by Tongtong Liang TongtongLiang/sihc-research-checkpoints
Selected Group 1 epoch-200 model weights (manifest.json), plus historical flagship and scaling-law weights (manifestflagshipscaling.json). For the large-model batch, epoch = step / 1251. Historical 40x1024 is epoch440; 28x1152 is epoch80.
Parameters—
Context—
Weights287.5 GB
License—
AccessOpen weights
Monthly Downloads—
Model Card
Selected Group 1 epoch-200 model weights (manifest.json), plus historical flagship and scaling-law weights (manifestflagshipscaling.json). For the large-model batch, epoch = step / 1251. Historical 40x1024 is epoch440; 28x1152 is epoch80. L-wide/deep are epoch200; H no-dropout includes epoch200 and epoch280; H projdrop0.2 is epoch80. Latest is not a claim of best FID. The H drop0.1 attempt has no checkpoint and is not included. Public derivatives retain raw model, EMA(s), and allowlisted architecture configuration only. No optimizer, RNG, credentials, W&B identity or local paths are included. These are inference weights, not full training-resume backups. Original checkpoints remain…
Excerpt from the card by Tongtong Liang.
Identity and Version
- Repository
- TongtongLiang/sihc-research-checkpoints
- Publisher
- Tongtong Liang
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- d4601d0fc44f7f552fd446376707c7f733149d3d
- First published
- 2026-09-17
- Last updated
- 2026-09-18
Files and Weights
103 files, 287.5 GB in total. The weights are 61 files totalling 287.5 GB in npz, pt.
Weights61 files · 287.5 GB
Configuration38 files · 83.7 KB
Documentation3 files · 2.5 KB
Repository1 file · 1.5 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| assets/fid_stats/jit_in256_stats.npz | Weights | 33.6 MB | 412046720c0d |
| flagship/sihc_5x8_d1024_velocity_repa_dinov3l/step_00550440.pt | Weights | 9.2 GB | 999056be2c1a |
| flagship/sihc_8x12x8_d1152_velocity_repa_dinov3l/step_00100080.pt | Weights | 8.2 GB | a35561e1a752 |
| group1/g1_fraction_hc_b16_velocity_p4_seed0/epoch_0200.pt | Weights | 1.0 GB | 82cf148d70a0 |
| group1/g1_hyperdit_b_p16p8_velocity_seed0/step_00250200.pt | Weights | 1.6 GB | a63491938825 |
| group1/g1_jit_b16_clean_p16_seed0/epoch_0200.pt | Weights | 1.1 GB | d12572b2feec |
| group1/g1_jit_b16_velocity_p16_seed0/epoch_0200.pt | Weights | 1.1 GB | 155ff87a57c5 |
| group1/g1_mhc4_b16_clean_p16_seed0/step_00250200.pt | Weights | 1.1 GB | 931a147184fe |
| group1/g1_mhc4_b16_velocity_p16_seed0/epoch_0200.pt | Weights | 1.1 GB | 16f9adea232e |
| group1/g1_mhc4_b16_velocity_p8_seed0/epoch_0200.pt | Weights | 1.1 GB | bef30ed80353 |
| group1/g1_sihc_b16_velocity_p4_ctx32s4_seed0/step_00250200.pt | Weights | 1.6 GB | 5cfe4d83670a |
| group1/g1_sihc_b16_velocity_p4_factorized_seed0/step_00250200.pt | Weights | 1.6 GB | 25a2ea28e002 |
| group1/g1_sihc_b16_velocity_p4_seed0/step_00250200.pt | Weights | 1.6 GB | abf2021b52f1 |
| group1/g1_sihc_b16_velocity_p8_seed0/step_00250200.pt | Weights | 1.6 GB | 3f7229c9e4ca |
| group1/g1_sihc_sublayer_b16_velocity_p4_ctx32s4_gmuon_seed0/step_00250200.pt | Weights | 1.6 GB | ef5d69dc653e |
| group1/g1_sihc_sublayer_b16_velocity_p4_ctx32s4_seed0/step_00250200.pt | Weights | 1.6 GB | 92449e253c54 |
| group1/g1_sihc_sublayer_b16_velocity_p4_seed0/step_00250200.pt | Weights | 1.6 GB | 65987de7724f |
| group1/g1_sihc_sublayer_b16_velocity_p8_seed0/step_00250200.pt | Weights | 1.6 GB | 71d5af366818 |
| group1/g1_sihc_sublayer_b_n16_rank48_meanout_velocity_seed0/step_00250200.pt | Weights | 1.6 GB | 257718d45129 |
| group1/g1_sihc_write_only_b16_velocity_p4_seed0/step_00250200.pt | Weights | 1.6 GB | 48cac4e4626b |
| group1/g1_static_hc4_b16_velocity_p8_seed0/step_00250200.pt | Weights | 1.6 GB | d6e6a1f0aa29 |
| raev2/imagenet_dinov3l_k7_latent_sihc_p8_d32_h1024_n4_seed42/ep-0000080.pt | Weights | 5.0 GB | caa853252047 |
| resume/external_group1/bsize_deco_b16_velocity_seed0/step_00250200.pt | Weights | 2.7 GB | 63a20f13ebc1 |
| resume/external_group1/bsize_dip_b16_velocity_seed0/step_00250200.pt | Weights | 2.7 GB | 56ef29bf8454 |
| resume/external_group1/bsize_pixeldit_b16_velocity_seed0/step_00250200.pt | Weights | 3.7 GB | a5f9a51cc73b |
| resume/flagship/sihc_5x8_d1024_velocity_repa_dinov3l/step_00550440.pt | Weights | 15.4 GB | 93736d3ba5a4 |
| resume/flagship/sihc_8x12x8_d1152_velocity_repa_dinov3l/step_00100080.pt | Weights | 13.7 GB | 213e53174427 |
| resume/group1/g1_fraction_hc_b16_velocity_p4_seed0/epoch_0200.pt | Weights | 2.1 GB | 7078bedd9779 |
| resume/group1/g1_hyperdit_b_p16p8_velocity_seed0/step_00250200.pt | Weights | 2.6 GB | bd467854a78a |
| resume/group1/g1_jit_b16_clean_p16_seed0/epoch_0200.pt | Weights | 2.1 GB | 8b22e837596c |
| resume/group1/g1_jit_b16_velocity_p16_seed0/epoch_0200.pt | Weights | 2.1 GB | 515e78e8d95e |
| resume/group1/g1_mhc4_b16_clean_p16_seed0/step_00250200.pt | Weights | 2.1 GB | 477aa7506378 |
| resume/group1/g1_mhc4_b16_velocity_p16_seed0/epoch_0200.pt | Weights | 2.1 GB | ef42600e77d7 |
| resume/group1/g1_mhc4_b16_velocity_p8_seed0/epoch_0200.pt | Weights | 2.1 GB | 56dd983a372a |
| resume/group1/g1_sihc_b16_velocity_p4_ctx32s4_seed0/step_00250200.pt | Weights | 2.6 GB | e1f365a85851 |
| resume/group1/g1_sihc_b16_velocity_p4_factorized_seed0/step_00250200.pt | Weights | 2.6 GB | 75595f7b6c83 |
| resume/group1/g1_sihc_b16_velocity_p4_seed0/step_00250200.pt | Weights | 2.6 GB | a467b4a748c0 |
| resume/group1/g1_sihc_b16_velocity_p8_seed0/step_00250200.pt | Weights | 2.6 GB | 793112c14bcf |
| resume/group1/g1_sihc_sublayer_b16_velocity_p4_ctx32s4_gmuon_seed0/step_00250200.pt | Weights | 2.1 GB | c2a8c2b5c62b |
| resume/group1/g1_sihc_sublayer_b16_velocity_p4_ctx32s4_seed0/step_00250200.pt | Weights | 2.6 GB | 3278712789da |
| resume/group1/g1_sihc_sublayer_b16_velocity_p4_seed0/step_00250200.pt | Weights | 2.6 GB | 4bdd5ac5b8a7 |
| resume/group1/g1_sihc_sublayer_b16_velocity_p8_seed0/step_00250200.pt | Weights | 2.6 GB | 6feef8f619bd |
| resume/group1/g1_sihc_sublayer_b_n16_rank48_meanout_velocity_seed0/step_00250200.pt | Weights | 2.7 GB | 185b339ff9df |
| resume/group1/g1_sihc_write_only_b16_velocity_p4_seed0/step_00250200.pt | Weights | 2.6 GB | df09e1787810 |
| resume/group1/g1_static_hc4_b16_velocity_p8_seed0/step_00250200.pt | Weights | 2.6 GB | 54de36e48b46 |
| resume/group6/g6_jit_b16_whiten_clean_seed0/step_00250200.pt | Weights | 2.6 GB | b92ab66a75f7 |
| resume/group6/g6_jit_b16_whiten_velocity_seed0/step_00250200.pt | Weights | 2.6 GB | 03b23bbad3cb |
| resume/raev2/imagenet_dinov3l_k7_latent_sihc_p8_d32_h1024_n4_ep25_decay1e4hold_seed42/ep-0000080.pt | Weights | 7.5 GB | f76bebf1a7da |
| resume/raev2/imagenet_dinov3l_k7_latent_sihc_p8_d32_h1024_n4_seed42/ep-0000080.pt | Weights | 7.5 GB | d9a6dd211d96 |
| resume/raev2/imagenet_dinov3l_k7_latent_sihc_p8_d32_h1024_n4_slotmod_fused8_origlr_bf16_unguided_e20_seed42/ep-0000080.pt | Weights | 7.5 GB | 80fde2058377 |
| resume/raev2/sihc_d32_direct4_ctx32s8_concatrms_e80_seed42/ep-0000080.pt | Weights | 7.5 GB | d0c0dd051acf |
| resume/scaling_law/g1_sihc_h_ctx32s8_velocity_norepa_mlp_recompute_projdrop02_seed0/step_00100080.pt | Weights | 19.0 GB | 8035653f1489 |
| resume/scaling_law/g1_sihc_h_ctx32s8_velocity_norepa_mlp_recompute_seed0/step_00250200.pt | Weights | 19.0 GB | ec21b1e6b9c1 |
| resume/scaling_law/g1_sihc_h_ctx32s8_velocity_norepa_mlp_recompute_seed0/step_00350280.pt | Weights | 19.0 GB | 143faa213b6c |
| resume/scaling_law/g1_sihc_l_deep_ctx32s12_velocity_norepa_seed0/step_00250200.pt | Weights | 10.3 GB | a7584056dcef |
| resume/scaling_law/g1_sihc_l_wide_ctx32s8_velocity_norepa_seed0/step_00250200.pt | Weights | 9.2 GB | 4f209ae0ed2f |
| scaling_law/g1_sihc_h_ctx32s8_velocity_norepa_mlp_recompute_projdrop02_seed0/step_00100080.pt | Weights | 11.4 GB | 235fcf240d38 |
| scaling_law/g1_sihc_h_ctx32s8_velocity_norepa_mlp_recompute_seed0/step_00250200.pt | Weights | 11.4 GB | 415e56784600 |
| scaling_law/g1_sihc_h_ctx32s8_velocity_norepa_mlp_recompute_seed0/step_00350280.pt | Weights | 11.4 GB | c21bf006e2ff |
| scaling_law/g1_sihc_l_deep_ctx32s12_velocity_norepa_seed0/step_00250200.pt | Weights | 6.2 GB | f865cc248ca4 |
| scaling_law/g1_sihc_l_wide_ctx32s8_velocity_norepa_seed0/step_00250200.pt | Weights | 5.5 GB | 22b63982e81c |
| manifest.json | Configuration | 6.6 KB | — |
| manifest_flagship_scaling.json | Configuration | 2.8 KB | — |
| manifest_raev2_e80.json | Configuration | 583 B | — |
| manifest_resume.json | Configuration | 18.5 KB | — |
| raev2/imagenet_dinov3l_k7_latent_sihc_p8_d32_h1024_n4_seed42/model_config.json | Configuration | 1.7 KB | — |
| resume/external_group1/bsize_deco_b16_velocity_seed0/resume_config.json | Configuration | 609 B | — |
| resume/external_group1/bsize_dip_b16_velocity_seed0/resume_config.json | Configuration | 607 B | — |
| resume/external_group1/bsize_pixeldit_b16_velocity_seed0/resume_config.json | Configuration | 617 B | — |
| resume/flagship/sihc_5x8_d1024_velocity_repa_dinov3l/resume_config.json | Configuration | 1.1 KB | — |
| resume/flagship/sihc_8x12x8_d1152_velocity_repa_dinov3l/resume_config.json | Configuration | 1.1 KB | — |
| resume/group1/g1_fraction_hc_b16_velocity_p4_seed0/resume_config.json | Configuration | 1.3 KB | — |
| resume/group1/g1_hyperdit_b_p16p8_velocity_seed0/resume_config.json | Configuration | 1.6 KB | — |
| resume/group1/g1_jit_b16_clean_p16_seed0/resume_config.json | Configuration | 1.2 KB | — |
| resume/group1/g1_jit_b16_velocity_p16_seed0/resume_config.json | Configuration | 1.2 KB | — |
| resume/group1/g1_mhc4_b16_clean_p16_seed0/resume_config.json | Configuration | 1.4 KB | — |
| resume/group1/g1_mhc4_b16_velocity_p16_seed0/resume_config.json | Configuration | 1.2 KB | — |
| resume/group1/g1_mhc4_b16_velocity_p8_seed0/resume_config.json | Configuration | 1.3 KB | — |
| resume/group1/g1_sihc_b16_velocity_p4_ctx32s4_seed0/resume_config.json | Configuration | 1.3 KB | — |
| resume/group1/g1_sihc_b16_velocity_p4_factorized_seed0/resume_config.json | Configuration | 1.1 KB | — |
| resume/group1/g1_sihc_b16_velocity_p4_seed0/resume_config.json | Configuration | 1.1 KB | — |
| resume/group1/g1_sihc_b16_velocity_p8_seed0/resume_config.json | Configuration | 1.1 KB | — |
| resume/group1/g1_sihc_sublayer_b16_velocity_p4_ctx32s4_gmuon_seed0/resume_config.json | Configuration | 1.8 KB | — |
| resume/group1/g1_sihc_sublayer_b16_velocity_p4_ctx32s4_seed0/resume_config.json | Configuration | 1.6 KB | — |
| resume/group1/g1_sihc_sublayer_b16_velocity_p4_seed0/resume_config.json | Configuration | 1.6 KB | — |
| resume/group1/g1_sihc_sublayer_b16_velocity_p8_seed0/resume_config.json | Configuration | 1.6 KB | — |
| resume/group1/g1_sihc_sublayer_b_n16_rank48_meanout_velocity_seed0/resume_config.json | Configuration | 1.6 KB | — |
| resume/group1/g1_sihc_write_only_b16_velocity_p4_seed0/resume_config.json | Configuration | 1.1 KB | — |
| resume/group1/g1_static_hc4_b16_velocity_p8_seed0/resume_config.json | Configuration | 1.6 KB | — |
| resume/group6/g6_jit_b16_whiten_clean_seed0/resume_config.json | Configuration | 1.6 KB | — |
| resume/group6/g6_jit_b16_whiten_velocity_seed0/resume_config.json | Configuration | 1.6 KB | — |
| resume/raev2/imagenet_dinov3l_k7_latent_sihc_p8_d32_h1024_n4_ep25_decay1e4hold_seed42/resume_config.json | Configuration | 2.8 KB | — |
| resume/raev2/imagenet_dinov3l_k7_latent_sihc_p8_d32_h1024_n4_seed42/resume_config.json | Configuration | 4.0 KB | — |
| resume/raev2/imagenet_dinov3l_k7_latent_sihc_p8_d32_h1024_n4_slotmod_fused8_origlr_bf16_unguided_e20_seed42/resume_config.json | Configuration | 4.0 KB | — |
| resume/raev2/sihc_d32_direct4_ctx32s8_concatrms_e80_seed42/resume_config.json | Configuration | 4.0 KB | — |
| resume/scaling_law/g1_sihc_h_ctx32s8_velocity_norepa_mlp_recompute_projdrop02_seed0/resume_config.json | Configuration | 1.7 KB | — |
| resume/scaling_law/g1_sihc_h_ctx32s8_velocity_norepa_mlp_recompute_seed0/resume_config.json | Configuration | 1.7 KB | — |
| resume/scaling_law/g1_sihc_l_deep_ctx32s12_velocity_norepa_seed0/resume_config.json | Configuration | 1.7 KB | — |
| resume/scaling_law/g1_sihc_l_wide_ctx32s8_velocity_norepa_seed0/resume_config.json | Configuration | 1.5 KB | — |
| README.md | Documentation | 1.1 KB | — |
| assets/fid_stats/README.md | Documentation | 364 B | — |
| resume/README.md | Documentation | 970 B | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 287.5 GB
Download from Tongtong Liang
Released by Tongtong Liang through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 287.5 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.