Hugging Face's logo

FastVideo
/
FastVideo-FastH3-Trim-8-Step

Text-to-Video
Diffusers
Safetensors
MiniMaxH3ModularPipeline
video-generation
audio
fastvideo
bf16
pruned
Model card Files
xet
 Community
aryan5v commited on
Commit
4021956
·
verified ·
1 Parent(s): 9c30793

Upload metadata.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. metadata.json +218 -0
metadata.json ADDED
@@ -0,0 +1,218 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model_dir": "/mnt/nfs/vlm-aryan/fasth3-14b-2step-qad-20260829/experiments/s42-r16-dmd8-vsa80-20260929-v1/initial-student",
3
+ "config": {
4
+ "callbacks": {
5
+ "grad_clip": {
6
+ "_target_": "fastvideo.train.callbacks.grad_clip.GradNormClipCallback",
7
+ "max_grad_norm": 1.0
8
+ },
9
+ "validation": {
10
+ "_target_": "fastvideo.train.callbacks.validation.ValidationCallback",
11
+ "dataset_file": "/mnt/nfs/vlm-aryan/fasth3-14b-2step-qad-20260829/experiments/s42-r16-dmd8-vsa80-20260929-v1/validation/val8.json",
12
+ "every_steps": 100,
13
+ "guidance_scale": 1.0,
14
+ "max_record_num_frames": 345,
15
+ "num_videos_per_prompt": 1,
16
+ "offload_training_state": true,
17
+ "pipeline_target": "fastvideo.pipelines.basic.minimax_h3.minimax_h3_pipeline.MiniMaxH3Pipeline",
18
+ "run_at_start": true,
19
+ "sampling_steps": [
20
+ 8
21
+ ],
22
+ "sampling_timesteps": [
23
+ 999,
24
+ 874,
25
+ 749,
26
+ 624,
27
+ 500,
28
+ 375,
29
+ 250,
30
+ 125
31
+ ],
32
+ "text_encoder_cpu_offload": true,
33
+ "use_record_dimensions": true,
34
+ "use_validation_media_conditioning": false,
35
+ "vae_cpu_offload": true
36
+ }
37
+ },
38
+ "method": {
39
+ "_target_": "fastvideo.train.methods.distribution_matching.dmd2.DMD2Method",
40
+ "audio_anchor": {
41
+ "weight": 0.1
42
+ },
43
+ "cfg_uncond": {
44
+ "text": "zero"
45
+ },
46
+ "dmd_denoising_steps": [
47
+ 999,
48
+ 874,
49
+ 749,
50
+ 624,
51
+ 500,
52
+ 375,
53
+ 250,
54
+ 125
55
+ ],
56
+ "fake_score_betas": [
57
+ 0.9,
58
+ 0.999
59
+ ],
60
+ "fake_score_learning_rate": "2e-06",
61
+ "fake_score_loss_space": "x0",
62
+ "fake_score_lr_scheduler": "constant",
63
+ "generator_update_interval": 5,
64
+ "max_timestep_ratio": 0.999,
65
+ "min_timestep_ratio": 0.001,
66
+ "modality_dmd_safeguards": {
67
+ "audio": {
68
+ "denom_floor_ratio": 0.05,
69
+ "grad_cap": 100.0
70
+ }
71
+ },
72
+ "real_score_guidance_scale": 1.0,
73
+ "rollout_carry": true,
74
+ "rollout_carry_slots": 8,
75
+ "rollout_mode": "simulate",
76
+ "rollout_sample_type": "ode",
77
+ "score_timestep_continuous": true,
78
+ "score_timestep_shift": 2.4,
79
+ "score_timestep_warp_max": 0.999
80
+ },
81
+ "models": {
82
+ "critic": {
83
+ "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3DMDModel",
84
+ "attention_backend": "FLASH_ATTN",
85
+ "disable_custom_init_weights": true,
86
+ "enable_gradient_checkpointing_type": "full",
87
+ "init_from": "/mnt/nfs/vlm-aryan/fasth3-14b-2step-qad-20260829/release-candidates/base-h3-teacher-complete-v1",
88
+ "trainable": true
89
+ },
90
+ "student": {
91
+ "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3DMDModel",
92
+ "attention_backend": "VIDEO_SPARSE_ATTN_H3",
93
+ "enable_gradient_checkpointing_type": "full",
94
+ "init_from": "/mnt/nfs/vlm-aryan/fasth3-14b-2step-qad-20260829/experiments/s42-r16-dmd8-vsa80-20260929-v1/initial-student",
95
+ "trainable": true
96
+ },
97
+ "teacher": {
98
+ "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3DMDModel",
99
+ "attention_backend": "FLASH_ATTN",
100
+ "disable_custom_init_weights": true,
101
+ "init_from": "/mnt/nfs/vlm-aryan/fasth3-14b-2step-qad-20260829/release-candidates/base-h3-teacher-complete-v1",
102
+ "trainable": false
103
+ }
104
+ },
105
+ "pipeline": {
106
+ "audio_scheduler_shift": 3.0,
107
+ "dit_config": {
108
+ "uniform_parameter_dtype": false
109
+ },
110
+ "video_scheduler_shift": 10.0
111
+ },
112
+ "training": {
113
+ "checkpoint": {
114
+ "checkpointing_start_step": 100,
115
+ "checkpoints_total_limit": 0,
116
+ "inference_checkpoint_dtype": "bfloat16",
117
+ "inference_checkpoint_role": "student",
118
+ "output_dir": "/mnt/nfs/vlm-aryan/fasth3-14b-2step-qad-20260829/experiments/s42-r16-dmd8-vsa80-20260929-v1/runs/vsa80-32gpu-anchor-v2",
119
+ "require_complete_training_checkpoint": true,
120
+ "resume_from_checkpoint": "latest",
121
+ "save_inference_checkpoint_on_validation": true,
122
+ "training_state_checkpointing_steps": 100
123
+ },
124
+ "data": {
125
+ "data_path": [
126
+ "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_video_nuva_50k_720_mixed_len/data",
127
+ "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_video_5s_768p/data",
128
+ "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_video_nuva_10k_720_mixed_len/data",
129
+ "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_video_nuva_10k_mixed_res_len/data",
130
+ "/mnt/lustre/vlm-shared/h3_t2av_preprocessed/v10_mixed_native_v3/h3_t2av_fastgen_vidprom_150k/data"
131
+ ],
132
+ "dataloader_num_workers": 0,
133
+ "native_shape_bucketing": true,
134
+ "num_frames": 124,
135
+ "num_height": 768,
136
+ "num_latent_t": 37,
137
+ "num_width": 1344,
138
+ "preprocessed_data_type": "t2va",
139
+ "seed": 42,
140
+ "train_batch_size": 1,
141
+ "training_cfg_rate": 0.0
142
+ },
143
+ "distributed": {
144
+ "hsdp_replicate_dim": 1,
145
+ "hsdp_shard_dim": 32,
146
+ "num_gpus": 32,
147
+ "sp_size": 4,
148
+ "tp_size": 1
149
+ },
150
+ "dit_precision": "fp32",
151
+ "loop": {
152
+ "gradient_accumulation_steps": 8,
153
+ "max_train_steps": 1500
154
+ },
155
+ "model": {
156
+ "enable_gradient_checkpointing_type": "full",
157
+ "enable_torch_compile": true,
158
+ "precondition_outputs": false,
159
+ "torch_compile_kwargs": {
160
+ "dynamic": true,
161
+ "recompile_limit": 32
162
+ }
163
+ },
164
+ "optimizer": {
165
+ "betas": [
166
+ 0.9,
167
+ 0.999
168
+ ],
169
+ "learning_rate": "2e-06",
170
+ "lr_scheduler": "constant",
171
+ "lr_warmup_steps": 0,
172
+ "weight_decay": 0.01
173
+ },
174
+ "tracker": {
175
+ "entity": "aryan5v-san-jose-state-university",
176
+ "project_name": "compacth3-s42-r16-dmd8-20260930",
177
+ "run_name": "new42-r16-vsa80-shift10-32gpu-audio-anchor-w0.1-v2",
178
+ "trackers": [
179
+ "wandb"
180
+ ]
181
+ },
182
+ "vsa": {
183
+ "sparsity": 0.8,
184
+ "tile_size": 64
185
+ }
186
+ }
187
+ },
188
+ "dtype": "bfloat16",
189
+ "format_version": 1,
190
+ "kind": "inference",
191
+ "max_shard_size_bytes": 5368709120,
192
+ "module": "transformer",
193
+ "role": "student",
194
+ "shard_count": 8,
195
+ "shard_file_sizes": [
196
+ 5364257712,
197
+ 5203335128,
198
+ 5126264664,
199
+ 5126264680,
200
+ 5126264664,
201
+ 5126264672,
202
+ 5126264624,
203
+ 1316821384
204
+ ],
205
+ "shard_sizes": [
206
+ 5364246528,
207
+ 5203325952,
208
+ 5126255616,
209
+ 5126255616,
210
+ 5126255616,
211
+ 5126255616,
212
+ 5126255616,
213
+ 1316819456
214
+ ],
215
+ "step": 300,
216
+ "tensor_count": 585,
217
+ "total_size": 37515670016
218
+ }