stage

2025-03-04 00:51:21 +00:00
parent 1e08893546
commit 31bbdcd2a4
12 changed files with 611 additions and 80 deletions
@@ -1,32 +1,33 @@
 #!/bin/bash

-num_gpus=4
-export MODEL_BASE=data/FastHunyuan
-torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29503 \
-    fastvideo/sample/sample_t2v_hunyuan.py \
-    --height 720 \
-    --width 1280 \
-    --num_frames 125 \
-    --num_inference_steps 6 \
-    --guidance_scale 1 \
-    --embedded_cfg_scale 6 \
-    --flow_shift 17 \
-    --flow-reverse \
-    --prompt ./assets/prompt.txt \
-    --seed 1024 \
-    --output_path outputs_video/hunyuan/vae_sp/ \
-    --model_path $MODEL_BASE \
-    --dit-weight ${MODEL_BASE}/hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states.pt \
-    --vae-sp
+# num_gpus=4
+# export MODEL_BASE=data/FastHunyuan
+# torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29503 \
+#     fastvideo/sample/sample_t2v_hunyuan.py \
+#     --height 720 \
+#     --width 1280 \
+#     --num_frames 125 \
+#     --num_inference_steps 6 \
+#     --guidance_scale 1 \
+#     --embedded_cfg_scale 6 \
+#     --flow_shift 17 \
+#     --flow-reverse \
+#     --prompt ./assets/prompt.txt \
+#     --seed 1024 \
+#     --output_path outputs_video/hunyuan/vae_sp/ \
+#     --model_path $MODEL_BASE \
+#     --dit-weight ${MODEL_BASE}/hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states.pt \
+#     --vae-sp

 # Inference original Hunyuan
-num_gpus=4
+num_gpus=1
 export MODEL_BASE=data/hunyuan
+mask_strategy_file_path=assets/test_mask_strategy_hunyuan.json
 torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29503 \
    fastvideo/sample/sample_t2v_hunyuan.py \
-    --height 720 \
+    --height 768 \
    --width 1280 \
-    --num_frames 125 \
+    --num_frames 45 \
    --num_inference_steps 50 \
    --guidance_scale 1 \
    --embedded_cfg_scale 6 \
@@ -34,7 +35,11 @@ torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29503 \
    --flow-reverse \
    --prompt ./assets/prompt.txt \
    --seed 1024 \
-    --output_path outputs_video/hunyuan/vae_sp/ \
+    --output_path outputs_video/hunyuan/flex_attention_235/ \
    --model_path $MODEL_BASE \
-    --dit-weight ${MODEL_BASE}/hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states.pt \
-    --vae-sp
+    --mask_strategy_file_path $mask_strategy_file_path \
+    --dit-weight ${MODEL_BASE}/hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states_fp8.pt \
+    --vae-sp \
+    --use-fp8 \
+    --use-cpu-offload \
+    --vae_tiling
@@ -1,39 +1,39 @@
 #!/bin/bash

 # Inference with STA + Teacache
-num_gpus=1 # currently it should be a factor of 30
-mask_strategy_file_path=assets/mask_strategy_hunyuan.json
-export MODEL_BASE=data/hunyuan
-rel_l1_thresh=0.15
-CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29603 \
-    fastvideo/sample/sample_t2v_hunyuan_STA.py \
-    --height 768 \
-    --width 1280 \
-    --num_frames 117 \
-    --num_inference_steps 50 \
-    --guidance_scale 1 \
-    --embedded_cfg_scale 6 \
-    --flow_shift 7 \
-    --flow-reverse \
-    --prompt ./assets/prompt.txt \
-    --seed 12345 \
-    --output_path outputs_video/hunyuan_STA/ \
-    --model_path $MODEL_BASE \
-    --mask_strategy_file_path $mask_strategy_file_path \
-    --rel_l1_thresh $rel_l1_thresh \
-    --dit-weight ${MODEL_BASE}/hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states.pt \
-    --vae-sp \
-    --enable_teacache 
+# num_gpus=1 # currently it should be a factor of 30
+# mask_strategy_file_path=assets/mask_strategy_hunyuan.json
+# export MODEL_BASE=data/hunyuan
+# rel_l1_thresh=0.15
+# CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29603 \
+#     fastvideo/sample/sample_t2v_hunyuan_STA.py \
+#     --height 768 \
+#     --width 1280 \
+#     --num_frames 117 \
+#     --num_inference_steps 50 \
+#     --guidance_scale 1 \
+#     --embedded_cfg_scale 6 \
+#     --flow_shift 7 \
+#     --flow-reverse \
+#     --prompt ./assets/prompt.txt \
+#     --seed 12345 \
+#     --output_path outputs_video/hunyuan_STA/ \
+#     --model_path $MODEL_BASE \
+#     --mask_strategy_file_path $mask_strategy_file_path \
+#     --rel_l1_thresh $rel_l1_thresh \
+#     --dit-weight ${MODEL_BASE}/hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states.pt \
+#     --vae-sp \
+#     --enable_teacache 

 # Inference with STA only
 num_gpus=1
-mask_strategy_file_path=assets/mask_strategy_hunyuan.json
+mask_strategy_file_path=assets/test_mask_strategy_hunyuan.json
 export MODEL_BASE=data/hunyuan
-CUDA_VISIBLE_DEVICES=1 torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29603 \
+torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29603 \
    fastvideo/sample/sample_t2v_hunyuan_STA.py \
    --height 768 \
    --width 1280 \
-    --num_frames 117 \
+    --num_frames 45 \
    --num_inference_steps 50 \
    --guidance_scale 1 \
    --embedded_cfg_scale 6 \
@@ -41,9 +41,11 @@ CUDA_VISIBLE_DEVICES=1 torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_p
    --flow-reverse \
    --prompt ./assets/prompt.txt \
    --seed 12345 \
-    --output_path outputs_video/hunyuan_STA/ \
+    --output_path outputs_video/hunyuan_STA/test_4090/ \
    --model_path $MODEL_BASE \
    --mask_strategy_file_path $mask_strategy_file_path \
-    --dit-weight ${MODEL_BASE}/hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states.pt \
+    --dit-weight ${MODEL_BASE}/hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states_fp8.pt \
    --vae-sp \
-    --enable_torch_compile
+    --use-fp8 \
+    --use-cpu-offload \
+    --vae_tiling
@@ -1,6 +1,6 @@
 #!/bin/bash

-num_gpus=4
+num_gpus=1
 torchrun --nnodes=1 --nproc_per_node=$num_gpus --master_port 29503 \
    fastvideo/sample/sample_t2v_hunyuan_hf.py \
    --model_path ~/data/hunyuan_diffusers/ \