Skip to content

GitLab

Commit e2778d0d authored by litzh's avatar litzh
Browse files

Initial commit

parents
Pipeline #3370 canceled with stages
Hide whitespace changes
Inline Side-by-side
Showing with 680 additions and 0 deletions
configs/seko_talk/seko_talk_13_fp8_dist_bucket_shape_8gpus_5s_realtime.json 0 → 100755
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 360,
"audio_sr": 16000,
"target_video_length": 81,
"resize_mode": "adaptive",
"bucket_shape": {
"0.667": [
[
480,
832
],
[
544,
960
]
],
"1.500": [
[
832,
480
],
[
960,
544
]
],
"1.000": [
[
480,
480
],
[
576,
576
],
[
704,
704
]
]
},
"self_attn_1_type": "sage_attn2",
"cross_attn_1_type": "sage_attn2",
"cross_attn_2_type": "sage_attn2",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true,
"parallel": {
"seq_p_size": 8,
"seq_p_attn_type": "ulysses"
},
"dit_quantized": true,
"dit_quant_scheme": "fp8-sgl",
"adapter_quantized": true,
"adapter_quant_scheme": "fp8-sgl",
"t5_quantized": true,
"t5_quant_scheme": "fp8-sgl"
}
configs/seko_talk/seko_talk_14_fp8_dist_bucket_shape_8gpus_1s_realtime.json 0 → 100755
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 360,
"audio_sr": 16000,
"target_video_length": 17,
"prev_frame_length": 1,
"resize_mode": "adaptive",
"bucket_shape": {
"0.667": [
[
480,
832
],
[
544,
960
]
],
"1.500": [
[
832,
480
],
[
960,
544
]
],
"1.000": [
[
480,
480
],
[
576,
576
],
[
704,
704
]
]
},
"self_attn_1_type": "flash_attn3",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true,
"parallel": {
"seq_p_size": 8,
"seq_p_attn_type": "ulysses"
},
"dit_quantized": true,
"dit_quant_scheme": "fp8-sgl",
"adapter_quantized": true,
"adapter_quant_scheme": "fp8-sgl",
"t5_quantized": true,
"t5_quant_scheme": "fp8-sgl"
}
configs/seko_talk/seko_talk_15_base_compile.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 360,
"audio_sr": 16000,
"target_video_length": 81,
"resize_mode": "adaptive",
"self_attn_1_type": "flash_attn3",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true,
"compile": true,
"compile_shapes": [
[
480,
832
],
[
544,
960
],
[
720,
1280
],
[
832,
480
],
[
960,
544
],
[
1280,
720
],
[
480,
480
],
[
576,
576
],
[
704,
704
],
[
960,
960
]
]
}
configs/seko_talk/seko_talk_16_fp8_dist_compile.json 0 → 100755
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 360,
"audio_sr": 16000,
"target_video_length": 81,
"resize_mode": "adaptive",
"self_attn_1_type": "sage_attn2",
"cross_attn_1_type": "sage_attn2",
"cross_attn_2_type": "sage_attn2",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true,
"parallel": {
"seq_p_size": 8,
"seq_p_attn_type": "ulysses"
},
"clip_quantized": true,
"clip_quant_scheme": "fp8-sgl",
"dit_quantized": true,
"dit_quant_scheme": "fp8-sgl",
"adapter_quantized": true,
"adapter_quant_scheme": "fp8-sgl",
"t5_quantized": true,
"t5_quant_scheme": "fp8-sgl",
"compile": true,
"compile_shapes": [
[
480,
832
],
[
544,
960
],
[
720,
1280
],
[
832,
480
],
[
960,
544
],
[
1280,
720
],
[
480,
480
],
[
576,
576
],
[
704,
704
],
[
960,
960
]
]
}
configs/seko_talk/seko_talk_17_base_vsr.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 2,
"target_fps": 25,
"video_duration": 1,
"audio_sr": 16000,
"target_video_length": 25,
"resize_mode": "fixed_shape",
"fixed_shape": [
192,
320
],
"self_attn_1_type": "flash_attn3",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true,
"video_super_resolution": {
"scale": 2.0,
"seed": 0,
"model_path": "/base_code/FlashVSR/examples/WanVSR/FlashVSR"
}
}
configs/seko_talk/seko_talk_22_nbhd_attn.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 360,
"audio_sr": 16000,
"target_video_length": 81,
"resize_mode": "adaptive",
"self_attn_1_type": "nbhd_attn",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true
}
configs/seko_talk/seko_talk_23_fp8_dist_nbhd_attn.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 360,
"audio_sr": 16000,
"target_video_length": 81,
"resize_mode": "adaptive",
"self_attn_1_type": "nbhd_attn",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true,
"parallel": {
"seq_p_size": 8,
"seq_p_attn_type": "ulysses"
},
"clip_quantized": true,
"clip_quant_scheme": "fp8-sgl",
"dit_quantized": true,
"dit_quant_scheme": "fp8-sgl",
"adapter_quantized": true,
"adapter_quant_scheme": "fp8-sgl",
"t5_quantized": true,
"t5_quant_scheme": "fp8-sgl"
}
configs/seko_talk/seko_talk_24_fp8_dist_compile_nbhd_attn.json 0 → 100755
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 360,
"audio_sr": 16000,
"target_video_length": 81,
"resize_mode": "adaptive",
"self_attn_1_type": "nbhd_attn",
"nbhd_attn_setting": {
"coefficient": [1.0, 0.25, 0.056]
},
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true,
"parallel": {
"seq_p_size": 8,
"seq_p_attn_type": "ulysses"
},
"clip_quantized": true,
"clip_quant_scheme": "fp8-sgl",
"dit_quantized": true,
"dit_quant_scheme": "fp8-sgl",
"adapter_quantized": true,
"adapter_quant_scheme": "fp8-sgl",
"t5_quantized": true,
"t5_quant_scheme": "fp8-sgl",
"compile": true,
"compile_shapes": [
[
480,
832
],
[
544,
960
],
[
720,
1280
],
[
832,
480
],
[
960,
544
],
[
1280,
720
],
[
480,
480
],
[
576,
576
],
[
704,
704
],
[
960,
960
]
]
}
configs/seko_talk/seko_talk_25_int8_dist_fp8_comm.json 0 → 100755
View file @ e2778d0d
{
"infer_steps": 2,
"target_fps": 16,
"video_duration": 5,
"audio_sr": 16000,
"target_video_length": 81,
"self_attn_1_type": "sage_attn3",
"cross_attn_1_type": "sage_attn3",
"cross_attn_2_type": "sage_attn3",
"sample_guide_scale": 1,
"sample_shift": 5,
"enable_cfg": false,
"use_31_block": true,
"cpu_offload": false,
"offload_granularity": "block",
"offload_ratio": 1,
"t5_cpu_offload": true,
"t5_quantized": true,
"t5_quant_scheme": "int8-q8f",
"clip_cpu_offload": true,
"clip_quantized": false,
"audio_encoder_cpu_offload": true,
"audio_adapter_cpu_offload": true,
"adapter_quantized": true,
"adapter_quant_scheme": "int8-q8f",
"vae_cpu_offload": true,
"use_tiling_vae": false,
"dit_quantized": true,
"dit_quant_scheme": "int8-q8f",
"resize_mode": "fixed_shape",
"fixed_shape": [
832,
480
],
"parallel": {
"seq_p_size": 8,
"seq_p_fp8_comm": true,
"seq_p_head_parallel": true,
"seq_p_attn_type": "ulysses"
}
}
configs/seko_talk/seko_talk_26_fp8_dist_fp8_comm.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 20,
"audio_sr": 16000,
"target_video_length": 81,
"resize_mode": "fixed_shape",
"fixed_shape": [
832,
480
],
"self_attn_1_type": "sage_attn3",
"cross_attn_1_type": "sage_attn3",
"cross_attn_2_type": "sage_attn3",
"sample_guide_scale": 1,
"sample_shift": 5,
"enable_cfg": false,
"use_31_block": true,
"cpu_offload": false,
"offload_granularity": "block",
"offload_ratio": 1,
"t5_cpu_offload": true,
"t5_quantized": true,
"t5_quant_scheme": "fp8-sgl",
"clip_cpu_offload": true,
"clip_quantized": false,
"audio_encoder_cpu_offload": true,
"audio_adapter_cpu_offload": true,
"adapter_quantized": true,
"adapter_quant_scheme": "fp8-sgl",
"vae_cpu_offload": true,
"use_tiling_vae": false,
"dit_quantized": true,
"dit_quant_scheme": "fp8-sgl",
"parallel": {
"seq_p_size": 4,
"seq_p_attn_type": "ring",
"seq_p_fp8_comm": true,
"seq_p_head_parallel": true,
"seq_p_tensor_fusion": false
}
}
configs/seko_talk/seko_talk_28_f2v.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 4,
"target_fps": 16,
"video_duration": 12,
"audio_sr": 16000,
"target_video_length": 81,
"prev_frame_length": 1,
"resize_mode": "adaptive",
"self_attn_1_type": "flash_attn3",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1.0,
"sample_shift": 5,
"enable_cfg": false,
"cpu_offload": false,
"use_31_block": true,
"f2v_process": true,
"lora_configs": [
{
"path": "lightx2v_I2V_14B_480p_cfg_step_distill_rank32_bf16.safetensors",
"strength": 1.0
}
]
}
configs/seko_talk/shot/rs2v/main.json 0 → 100644
View file @ e2778d0d
{
"lightx2v_path": "",
"clip_configs": [
{
"name":"rs2v_clip",
"path": "configs/seko_talk/shot/rs2v/rs2v.json"
}
]
}
configs/seko_talk/shot/rs2v/rs2v.json 0 → 100644
View file @ e2778d0d
{
"model_cls": "seko_talk",
"task": "rs2v",
"model_path":"/data/temp/SekoTalk-v2.7_beta1-bf16-step4",
"infer_steps": 4,
"target_fps": 16,
"audio_sr": 16000,
"resize_mode": "adaptive",
"self_attn_1_type": "flash_attn3",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1,
"sample_shift": 5,
"enable_cfg": false,
"use_31_block": true,
"target_video_length": 81,
"prev_frame_length": 0
}
configs/seko_talk/shot/stream/f2v.json 0 → 100644
View file @ e2778d0d
{
"model_cls": "seko_talk",
"task": "s2v",
"model_path":"Wan2.1-i2V1202-Audio-14B-720P/",
"infer_steps": 4,
"target_fps": 16,
"audio_sr": 16000,
"resize_mode": "adaptive",
"self_attn_1_type": "flash_attn3",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1,
"sample_shift": 5,
"enable_cfg": false,
"use_31_block": true,
"rope_type": "torch",
"target_video_length": 33,
"prev_frame_length": 1,
"f2v_process": true,
"cpu_offload": true,
"offload_granularity": "block",
"t5_cpu_offload": true,
"clip_cpu_offload": false,
"vae_cpu_offload": false,
"offload_ratio": 1,
"use_tiling_vae": true,
"audio_encoder_cpu_offload": true,
"audio_adapter_cpu_offload": false,
"lora_configs": [
{
"path": "lightx2v_I2V_14B_480p_cfg_step_distill_rank32_bf16.safetensors",
"strength": 1.0
}
]
}
configs/seko_talk/shot/stream/main.json 0 → 100644
View file @ e2778d0d
{
"lightx2v_path": "",
"clip_configs": [
{
"name":"s2v_clip",
"path": "configs/seko_talk/shot/stream/s2v.json"
},
{
"name":"f2v_clip",
"path": "configs/seko_talk/shot/stream/f2v.json"
}
]
}
configs/seko_talk/shot/stream/s2v.json 0 → 100644
View file @ e2778d0d
{
"model_cls": "seko_talk",
"task": "s2v",
"model_path":"Wan2.1-R2V721-Audio-14B-720P/",
"infer_steps": 4,
"target_fps": 16,
"audio_sr": 16000,
"resize_mode": "adaptive",
"self_attn_1_type": "flash_attn3",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 1,
"sample_shift": 5,
"enable_cfg": false,
"use_31_block": true,
"rope_type": "torch",
"target_video_length": 33,
"prev_frame_length": 5,
"cpu_offload": true,
"offload_granularity": "block",
"t5_cpu_offload": true,
"clip_cpu_offload": false,
"vae_cpu_offload": false,
"offload_ratio": 1,
"use_tiling_vae": true,
"audio_encoder_cpu_offload": true,
"audio_adapter_cpu_offload": false
}
configs/self_forcing/wan_t2v_sf.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 50,
"target_video_length": 81,
"text_len": 512,
"target_height": 480,
"target_width": 832,
"self_attn_1_type": "flash_attn2",
"cross_attn_1_type": "flash_attn2",
"cross_attn_2_type": "flash_attn2",
"seed": 0,
"sample_guide_scale": 6,
"sample_shift": 8,
"enable_cfg": false,
"cpu_offload": false,
"sf_config": {
"sf_type": "dmd",
"local_attn_size": -1,
"shift": 5.0,
"num_frame_per_block": 3,
"num_transformer_blocks": 30,
"frame_seq_length": 1560,
"num_output_frames": 21,
"num_inference_steps": 1000,
"denoising_step_list": [1000.0000, 937.5000, 833.3333, 625.0000]
}
}
configs/sparse_attn/spas_sage_attn/wan_i2v.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 40,
"target_video_length": 81,
"target_height": 480,
"target_width": 832,
"self_attn_1_type": "sage_attn",
"cross_attn_1_type": "sage_attn",
"cross_attn_2_type": "sage_attn",
"sample_guide_scale": 5,
"sample_shift": 3,
"enable_cfg": true,
"cpu_offload": false
}
configs/sparse_attn/spas_sage_attn/wan_t2v.json 0 → 100644
View file @ e2778d0d
{
"infer_steps": 50,
"target_video_length": 81,
"text_len": 512,
"target_height": 480,
"target_width": 832,
"self_attn_1_type": "spas_sage_attn",
"cross_attn_1_type": "spas_sage_attn",
"cross_attn_2_type": "spas_sage_attn",
"sample_guide_scale": 6,
"sample_shift": 8,
"enable_cfg": true,
"cpu_offload": false
}
configs/video_frame_interpolation/wan_t2v.json 0 → 100755
View file @ e2778d0d
{
"infer_steps": 50,
"target_video_length": 81,
"text_len": 512,
"target_height": 480,
"target_width": 832,
"self_attn_1_type": "flash_attn3",
"cross_attn_1_type": "flash_attn3",
"cross_attn_2_type": "flash_attn3",
"sample_guide_scale": 6,
"sample_shift": 8,
"enable_cfg": true,
"cpu_offload": false,
"video_frame_interpolation": {
"algo": "rife",
"target_fps": 24,
"model_path": "/path to flownet.pkl"
}
}
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment