Skip to content

GitLab

"megatron/vscode:/vscode.git/clone" did not exist on "8b94a16092df9f22daef3d97faa47911c43b1894"
Unverified Commit 94e593ce authored by gushiqiao's avatar gushiqiao Committed by GitHub
Browse files

[Recon] remove mm_config and support load single safetensors file (#375)

parent 954df466
Show whitespace changes
Inline Side-by-side
Showing with 44 additions and 63 deletions
configs/bench/lightx2v_3.json
View file @ 94e593ce
...@@ -10,8 +10,7 @@ ...@@ -10,8 +10,7 @@
"sample_shift": 5, "sample_shift": 5,
"enable_cfg": true, "enable_cfg": true,
"cpu_offload": false, "cpu_offload": false,
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Sgl-ActVllm" "dit_quant_scheme": "fp8-sgl",
},
"use_tiling_vae": true "use_tiling_vae": true
} }
configs/bench/lightx2v_3_distill.json
View file @ 94e593ce
...@@ -16,8 +16,7 @@ ...@@ -16,8 +16,7 @@
500, 500,
250 250
], ],
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Sgl-ActVllm" "dit_quant_scheme": "fp8-sgl",
},
"use_tiling_vae": true "use_tiling_vae": true
} }
configs/bench/lightx2v_4.json
View file @ 94e593ce
...@@ -10,9 +10,8 @@ ...@@ -10,9 +10,8 @@
"sample_shift": 5, "sample_shift": 5,
"enable_cfg": true, "enable_cfg": true,
"cpu_offload": false, "cpu_offload": false,
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Sgl-ActVllm" "dit_quant_scheme": "fp8-sgl",
},
"feature_caching": "Tea", "feature_caching": "Tea",
"coefficients": [ "coefficients": [
[ [
......
configs/bench/lightx2v_5.json
View file @ 94e593ce
...@@ -13,8 +13,7 @@ ...@@ -13,8 +13,7 @@
"offload_granularity": "block", "offload_granularity": "block",
"offload_ratio": 0.8, "offload_ratio": 0.8,
"t5_cpu_offload": true, "t5_cpu_offload": true,
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "fp8-sgl",
},
"use_tiling_vae": true "use_tiling_vae": true
} }
configs/bench/lightx2v_5_distill.json
View file @ 94e593ce
...@@ -19,8 +19,7 @@ ...@@ -19,8 +19,7 @@
500, 500,
250 250
], ],
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "fp8-sgl",
},
"use_tiling_vae": true "use_tiling_vae": true
} }
configs/bench/lightx2v_6.json
View file @ 94e593ce
...@@ -13,8 +13,7 @@ ...@@ -13,8 +13,7 @@
"offload_granularity": "block", "offload_granularity": "block",
"offload_ratio": 0.8, "offload_ratio": 0.8,
"t5_cpu_offload": true, "t5_cpu_offload": true,
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "fp8-sgl",
},
"use_tiling_vae": true "use_tiling_vae": true
} }
configs/bench/lightx2v_6_distill.json
View file @ 94e593ce
...@@ -19,8 +19,7 @@ ...@@ -19,8 +19,7 @@
500, 500,
250 250
], ],
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "fp8-sgl",
},
"use_tiling_vae": true "use_tiling_vae": true
} }
configs/distill/wan_i2v_distill_4step_cfg.json
View file @ 94e593ce
...@@ -15,5 +15,12 @@ ...@@ -15,5 +15,12 @@
750, 750,
500, 500,
250 250
] ],
"mm_config": {
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Vllm"
},
"dit_quantized": true,
"dit_quant_scheme": "fp8-sgl",
"dit_quantized_ckpt": "/data/nvme0/gushiqiao/models/hf_lightx2v_models/models/wan2.1_i2v_480p_scaled_fp8_e4m3_lightx2v_4step.safetensors"
} }
configs/distill/wan_t2v_distill_4step_cfg.json
View file @ 94e593ce
...@@ -16,5 +16,6 @@ ...@@ -16,5 +16,6 @@
750, 750,
500, 500,
250 250
] ],
"dit_original_ckpt": "/data/nvme0/gushiqiao/models/hf_lightx2v_models/models/wan2.1_t2v_14b_lightx2v_4step.safetensors"
} }
configs/offload/block/wan_t2v_block.json
View file @ 94e593ce
...@@ -13,7 +13,6 @@ ...@@ -13,7 +13,6 @@
"cpu_offload": true, "cpu_offload": true,
"offload_granularity": "block", "offload_granularity": "block",
"t5_cpu_offload": true, "t5_cpu_offload": true,
"mm_config": { "dit_quantized": true,
"mm_type": "W-int8-channel-sym-A-int8-channel-sym-dynamic-Vllm" "dit_quant_scheme": "fp8-sgl"
}
} }
configs/offload/disk/wan_i2v_phase_lazy_load_480p.json
View file @ 94e593ce
...@@ -13,9 +13,8 @@ ...@@ -13,9 +13,8 @@
"offload_granularity": "phase", "offload_granularity": "phase",
"t5_offload_granularity": "block", "t5_offload_granularity": "block",
"dit_quantized_ckpt": "/path/to/dit_quant_model", "dit_quantized_ckpt": "/path/to/dit_quant_model",
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Vllm" "dit_quant_scheme": "fp8-vllm",
},
"t5_cpu_offload": true, "t5_cpu_offload": true,
"t5_quantized": true, "t5_quantized": true,
"t5_quantized_ckpt": "/path/to/models_t5_umt5-xxl-enc-fp8.pth", "t5_quantized_ckpt": "/path/to/models_t5_umt5-xxl-enc-fp8.pth",
......
configs/offload/disk/wan_i2v_phase_lazy_load_720p.json
View file @ 94e593ce
...@@ -13,9 +13,8 @@ ...@@ -13,9 +13,8 @@
"offload_granularity": "phase", "offload_granularity": "phase",
"t5_offload_granularity": "block", "t5_offload_granularity": "block",
"dit_quantized_ckpt": "/path/to/dit_quant_model", "dit_quantized_ckpt": "/path/to/dit_quant_model",
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Vllm" "dit_quant_scheme": "fp8-vllm",
},
"t5_cpu_offload": true, "t5_cpu_offload": true,
"t5_quantized": true, "t5_quantized": true,
"t5_quantized_ckpt": "/path/to/models_t5_umt5-xxl-enc-fp8.pth", "t5_quantized_ckpt": "/path/to/models_t5_umt5-xxl-enc-fp8.pth",
......
configs/offload/phase/wan_t2v_phase.json
View file @ 94e593ce
...@@ -14,9 +14,8 @@ ...@@ -14,9 +14,8 @@
"t5_cpu_offload": true, "t5_cpu_offload": true,
"offload_granularity": "phase", "offload_granularity": "phase",
"dit_quantized_ckpt": "/path/to/dit_int8", "dit_quantized_ckpt": "/path/to/dit_int8",
"mm_config": { "dit_quantized": true,
"mm_type": "W-int8-channel-sym-A-int8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "int8-q8f",
},
"use_tiny_vae": true, "use_tiny_vae": true,
"tiny_vae_path": "/x2v_models/taew2_1.pth", "tiny_vae_path": "/x2v_models/taew2_1.pth",
"t5_offload_granularity": "block" "t5_offload_granularity": "block"
......
configs/quantization/hunyuan_i2v.json deleted 100755 → 0
View file @ 954df466
{
"infer_steps": 20,
"target_video_length": 33,
"i2v_resolution": "720p",
"attention_type": "flash_attn3",
"dit_quantized_ckpt": "/path/to/int8/model",
"mm_config": {
"mm_type": "W-int8-channel-sym-A-int8-channel-sym-dynamic-Vllm"
}
}
configs/quantization/wan_i2v.json
View file @ 94e593ce
...@@ -11,7 +11,6 @@ ...@@ -11,7 +11,6 @@
"enable_cfg": true, "enable_cfg": true,
"cpu_offload": false, "cpu_offload": false,
"dit_quantized_ckpt": "/path/to/int8/model", "dit_quantized_ckpt": "/path/to/int8/model",
"mm_config": { "dit_quantized": true,
"mm_type": "W-int8-channel-sym-A-int8-channel-sym-dynamic-Vllm" "dit_quant_scheme": "int8-vllm"
}
} }
configs/quantization/wan_i2v_q8f.json
View file @ 94e593ce
...@@ -10,9 +10,8 @@ ...@@ -10,9 +10,8 @@
"sample_shift": 5, "sample_shift": 5,
"enable_cfg": false, "enable_cfg": false,
"cpu_offload": false, "cpu_offload": false,
"mm_config": { "dit_quantized": true,
"mm_type": "W-int8-channel-sym-A-int8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "int8-q8f",
},
"t5_quantized": true, "t5_quantized": true,
"t5_quant_scheme": "int8-q8f", "t5_quant_scheme": "int8-q8f",
"clip_quantized": true, "clip_quantized": true,
......
configs/quantization/wan_i2v_torchao.json
View file @ 94e593ce
...@@ -10,9 +10,8 @@ ...@@ -10,9 +10,8 @@
"sample_shift": 5, "sample_shift": 5,
"enable_cfg": false, "enable_cfg": false,
"cpu_offload": false, "cpu_offload": false,
"mm_config": { "dit_quantized": true,
"mm_type": "W-int8-channel-sym-A-int8-channel-sym-dynamic-Torchao" "dit_quant_scheme": "int8-torchao",
},
"t5_quantized": true, "t5_quantized": true,
"t5_quant_scheme": "int8-torchao", "t5_quant_scheme": "int8-torchao",
"clip_quantized": true, "clip_quantized": true,
......
configs/seko_talk/L40s/1gpu/seko_talk_fp8.json 100644 → 100755
View file @ 94e593ce
...@@ -14,9 +14,8 @@ ...@@ -14,9 +14,8 @@
"use_31_block": false, "use_31_block": false,
"t5_quantized": true, "t5_quantized": true,
"t5_quant_scheme": "fp8-q8f", "t5_quant_scheme": "fp8-q8f",
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "fp8-q8f",
},
"adapter_quantized": true, "adapter_quantized": true,
"adapter_quant_scheme": "fp8", "adapter_quant_scheme": "fp8",
"cpu_offload": false, "cpu_offload": false,
......
configs/seko_talk/L40s/2gpu/seko_talk_fp8.json 100644 → 100755
View file @ 94e593ce
...@@ -14,9 +14,8 @@ ...@@ -14,9 +14,8 @@
"use_31_block": false, "use_31_block": false,
"t5_quantized": true, "t5_quantized": true,
"t5_quant_scheme": "fp8-q8f", "t5_quant_scheme": "fp8-q8f",
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "fp8-q8f",
},
"adapter_quantized": true, "adapter_quantized": true,
"adapter_quant_scheme": "fp8", "adapter_quant_scheme": "fp8",
"cpu_offload": false, "cpu_offload": false,
......
configs/seko_talk/L40s/4gpu/seko_talk_fp8.json 100644 → 100755
View file @ 94e593ce
...@@ -14,9 +14,8 @@ ...@@ -14,9 +14,8 @@
"use_31_block": false, "use_31_block": false,
"t5_quantized": true, "t5_quantized": true,
"t5_quant_scheme": "fp8-q8f", "t5_quant_scheme": "fp8-q8f",
"mm_config": { "dit_quantized": true,
"mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Q8F" "dit_quant_scheme": "fp8-q8f",
},
"adapter_quantized": true, "adapter_quantized": true,
"adapter_quant_scheme": "fp8", "adapter_quant_scheme": "fp8",
"cpu_offload": false, "cpu_offload": false,
......
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