checkpoint_metadata.json
6.0 KB · 205 lines · json Raw
1 {
2 "base_model_dir": "/mnt/lustre/vlm-k1kong/models/MiniMax-H3",
3 "config": {
4 "callbacks": {
5 "grad_clip": {
6 "_target_": "fastvideo.train.callbacks.grad_clip.GradNormClipCallback",
7 "max_grad_norm": 1.0
8 },
9 "latent_vis": {
10 "_target_": "fastvideo.train.callbacks.latent_vis.LatentVisCallback",
11 "every_steps": 100,
12 "fps": 24
13 },
14 "validation": {
15 "_target_": "fastvideo.train.callbacks.validation.ValidationCallback",
16 "dataset_file": "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/validation/heldout60.json",
17 "every_steps": 100,
18 "guidance_scale": 1.0,
19 "max_record_num_frames": 345,
20 "num_videos_per_prompt": 1,
21 "offload_training_state": true,
22 "pipeline_target": "fastvideo.pipelines.basic.minimax_h3.minimax_h3_pipeline.MiniMaxH3Pipeline",
23 "run_at_start": true,
24 "sampling_steps": [
25 4
26 ],
27 "text_encoder_cpu_offload": true,
28 "use_record_dimensions": true,
29 "use_validation_media_conditioning": false,
30 "vae_cpu_offload": true
31 }
32 },
33 "method": {
34 "_target_": "fastvideo.train.methods.distribution_matching.dmd2.DMD2Method",
35 "cfg_uncond": {
36 "text": "zero"
37 },
38 "dmd_denoising_steps": [
39 999,
40 749,
41 500,
42 250
43 ],
44 "fake_score_betas": [
45 0.9,
46 0.999
47 ],
48 "fake_score_learning_rate": 2e-06,
49 "fake_score_loss_space": "x0",
50 "fake_score_lr_scheduler": "constant",
51 "generator_update_interval": 5,
52 "max_timestep_ratio": 0.999,
53 "min_timestep_ratio": 0.001,
54 "real_score_guidance_scale": 1.0,
55 "rollout_carry": true,
56 "rollout_carry_slots": 8,
57 "rollout_mode": "simulate",
58 "rollout_sample_type": "ode",
59 "score_timestep_shift": 2.4
60 },
61 "models": {
62 "critic": {
63 "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3DMDModel",
64 "attention_backend": "FLASH_ATTN",
65 "disable_custom_init_weights": true,
66 "enable_gradient_checkpointing_type": "full",
67 "init_from": "/mnt/lustre/vlm-k1kong/models/MiniMax-H3",
68 "trainable": true
69 },
70 "student": {
71 "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3DMDModel",
72 "attention_backend": "VIDEO_SPARSE_ATTN_H3",
73 "enable_gradient_checkpointing_type": "full",
74 "init_from": "/mnt/lustre/vlm-k1kong/models/MiniMax-H3",
75 "trainable": true
76 },
77 "teacher": {
78 "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3DMDModel",
79 "attention_backend": "FLASH_ATTN",
80 "disable_custom_init_weights": true,
81 "init_from": "/mnt/lustre/vlm-k1kong/models/MiniMax-H3",
82 "trainable": false
83 }
84 },
85 "pipeline": {
86 "dit_config": {
87 "uniform_parameter_dtype": false
88 }
89 },
90 "training": {
91 "checkpoint": {
92 "checkpointing_start_step": 100,
93 "checkpoints_total_limit": 3,
94 "inference_checkpoint_dtype": "bfloat16",
95 "inference_checkpoint_role": "student",
96 "output_dir": "/mnt/lustre/vlm-wlsaidhi/fastvideo/outputs/minimax_h3_dmd2_sp4_fsdp32_v10p5_datafree_mixed_vsa64",
97 "require_complete_training_checkpoint": true,
98 "resume_from_checkpoint": "latest",
99 "save_inference_checkpoint_on_validation": true,
100 "training_state_checkpointing_steps": 100
101 },
102 "data": {
103 "data_path": [
104 "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_video_nuva_50k_720_mixed_len/data",
105 "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_video_5s_768p/data",
106 "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_video_nuva_10k_720_mixed_len/data",
107 "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_video_nuva_10k_mixed_res_len/data",
108 "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_fastgen_vidprom_150k/data"
109 ],
110 "dataloader_num_workers": 0,
111 "native_shape_bucketing": true,
112 "num_frames": 124,
113 "num_height": 768,
114 "num_latent_t": 37,
115 "num_width": 1344,
116 "preprocessed_data_type": "text_only",
117 "seed": 42,
118 "train_batch_size": 1,
119 "training_cfg_rate": 0.0
120 },
121 "distributed": {
122 "hsdp_replicate_dim": 1,
123 "hsdp_shard_dim": 32,
124 "num_gpus": 32,
125 "sp_size": 4,
126 "tp_size": 1
127 },
128 "dit_precision": "fp32",
129 "loop": {
130 "gradient_accumulation_steps": 8,
131 "max_train_steps": 4000
132 },
133 "model": {
134 "enable_gradient_checkpointing_type": "full",
135 "enable_torch_compile": true,
136 "precondition_outputs": false,
137 "torch_compile_kwargs": {
138 "dynamic": true,
139 "recompile_limit": 32
140 }
141 },
142 "optimizer": {
143 "betas": [
144 0.9,
145 0.999
146 ],
147 "learning_rate": 2e-06,
148 "lr_scheduler": "constant",
149 "lr_warmup_steps": 0,
150 "weight_decay": 0.01
151 },
152 "tracker": {
153 "project_name": "h3-dmd2-vsa",
154 "run_name": "dmd2_sp4_fsdp32_v10p5_datafree_mixed_vsa64"
155 },
156 "vsa": {
157 "sparsity": 0.9,
158 "tile_size": 64
159 }
160 }
161 },
162 "dtype": "bfloat16",
163 "format_version": 1,
164 "kind": "inference",
165 "max_shard_size_bytes": 5368709120,
166 "module": "transformer",
167 "role": "student",
168 "shard_count": 14,
169 "shard_file_sizes": [
170 5331008552,
171 5318721256,
172 5318721248,
173 4875566048,
174 5318505992,
175 5164580688,
176 5318721240,
177 5318721248,
178 4875566056,
179 5318505984,
180 5164580688,
181 5318721256,
182 5318721208,
183 2138941296
184 ],
185 "shard_sizes": [
186 5331000320,
187 5318715392,
188 5318715392,
189 4875560448,
190 5318500352,
191 5164574720,
192 5318715392,
193 5318715392,
194 4875560448,
195 5318500352,
196 5164574720,
197 5318715392,
198 5318715392,
199 2138938880
200 ],
201 "step": 1300,
202 "tensor_count": 688,
203 "total_size": 70099502592
204 }
205