olivame/unifolm-world-model-action
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

全链路 bf16 混合精度修正与 UNet FLOPS profiling

Browse Source 
- GroupNorm/LayerNorm bypass autocast，消除 bf16→fp32→bf16 转换开销
  - DDIM 调度系数 cast 到输入 dtype，attention mask 直接用 bf16 分配
  - alphas_cumprod 提升到 float64 保证数值精度
  - SinusoidalPosEmb 输出 dtype跟随模型精度
  - 新增 profile_unet.py 脚本及FLOPS 分析结果
  - 启用 TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL
  - case1 PSNR: 30.45 → 30.24（bf16 精度预期内波动）
This commit is contained in:
olivame
2026-02-08 16:01:30 +00:00
parent 75c798ded0
commit f86ab51a04
11 changed files with 464 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
unitree_z1_dual_arm_cleanup_pencils/case1/output.log
View File

@@ -1,12 +1,12 @@
2026-02-08 13:59:02.578826: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026-02-08 13:59:02.581891: I external/local_tsl/tsl/cuda/cudart_stub.cc:31] Could not find cuda drivers on your machine, GPU will not be used.
2026-02-08 13:59:02.613088: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2026-02-08 13:59:02.613125: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2026-02-08 13:59:02.614961: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2026-02-08 13:59:02.623180: I external/local_tsl/tsl/cuda/cudart_stub.cc:31] Could not find cuda drivers on your machine, GPU will not be used.
2026-02-08 13:59:02.623460: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
2026-02-08 15:47:30.035545: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026-02-08 15:47:30.038628: I external/local_tsl/tsl/cuda/cudart_stub.cc:31] Could not find cuda drivers on your machine, GPU will not be used.
2026-02-08 15:47:30.069635: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2026-02-08 15:47:30.069671: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2026-02-08 15:47:30.071534: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2026-02-08 15:47:30.080021: I external/local_tsl/tsl/cuda/cudart_stub.cc:31] Could not find cuda drivers on your machine, GPU will not be used.
2026-02-08 15:47:30.080300: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI AVX512_BF16 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2026-02-08 13:59:03.306638: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
2026-02-08 15:47:30.746161: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
[rank: 0] Global seed set to 123
/mnt/ASC1637/miniconda3/envs/unifolm-wma/lib/python3.10/site-packages/kornia/feature/lightglue.py:44: FutureWarning: `torch.cuda.amp.custom_fwd(args...)` is deprecated. Please use `torch.amp.custom_fwd(args..., device_type='cuda')` instead.
@torch.cuda.amp.custom_fwd(cast_inputs=torch.float32)
@@ -23,7 +23,7 @@ INFO:root:Loading pretrained ViT-H-14 weights (laion2b_s32b_b79k).
INFO:root:Loaded ViT-H-14 model config.
DEBUG:urllib3.connectionpool:https://hf-mirror.com:443 "HEAD /laion/CLIP-ViT-H-14-laion2B-s32B-b79K/resolve/main/open_clip_pytorch_model.bin HTTP/1.1" 302 0
INFO:root:Loading pretrained ViT-H-14 weights (laion2b_s32b_b79k).
/mnt/ASC1637/unifolm-world-model-action/scripts/evaluation/world_model_interaction.py:149: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
/mnt/ASC1637/unifolm-world-model-action/scripts/evaluation/world_model_interaction.py:183: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
state_dict = torch.load(ckpt, map_location="cpu")
>>> model checkpoint loaded.
>>> Load pre-trained model ...
@@ -36,6 +36,8 @@ INFO:root:Loading pretrained ViT-H-14 weights (laion2b_s32b_b79k).
✓ Projectors converted to bfloat16
✓ Encoders converted to bfloat16
✓ VAE converted to bfloat16
⚠ Found 601 fp32 params, converting to bf16
✓ All parameters converted to bfloat16
INFO:root:***** Configing Data *****
>>> unitree_z1_stackbox: 1 data samples loaded.
>>> unitree_z1_stackbox: data stats loaded.
@@ -60,10 +62,6 @@ DEBUG:PIL.PngImagePlugin:STREAM b'pHYs' 41 9
DEBUG:PIL.PngImagePlugin:STREAM b'IDAT' 62 4096
0%| | 0/8 [00:00<?, ?it/s]/mnt/ASC1637/miniconda3/envs/unifolm-wma/lib/python3.10/site-packages/torch/nn/functional.py:5501: UserWarning: Attempting to use hipBLASLt on an unsupported architecture! Overriding blas backend to hipblas (Triggered internally at ../aten/src/ATen/Context.cpp:296.)
proj = linear(q, w, b)
/mnt/ASC1637/miniconda3/envs/unifolm-wma/lib/python3.10/site-packages/torch/nn/functional.py:6278: UserWarning: Flash attention support on Navi31 GPU is still experimental. Enable it with TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1. (Triggered internally at ../aten/src/ATen/native/transformers/hip/sdp_utils.cpp:225.)
attn_output = scaled_dot_product_attention(
/mnt/ASC1637/miniconda3/envs/unifolm-wma/lib/python3.10/site-packages/torch/nn/functional.py:6278: UserWarning: Memory Efficient attention on Navi31 GPU is still experimental. Enable it with TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1. (Triggered internally at ../aten/src/ATen/native/transformers/hip/sdp_utils.cpp:269.)
proj = linear(q, w, b)
>>> Step 0: generating actions ...
>>> Step 0: interacting with world model ...
@@ -115,7 +113,7 @@ DEBUG:PIL.Image:Importing WmfImagePlugin
DEBUG:PIL.Image:Importing WmfImagePlugin
DEBUG:PIL.Image:Importing XbmImagePlugin
DEBUG:PIL.Image:Importing XpmImagePlugin
DEBUG:PIL.Image:Importing XVThumbImagePlugin
DEBUG:PIL.Image:Importing XVThumbImagePlugin
12%|█▎ | 1/8 [01:19<09:17, 79.58s/it]
25%|██▌ | 2/8 [02:38<07:54, 79.06s/it]
@@ -139,6 +137,6 @@ DEBUG:PIL.Image:Importing XVThumbImagePlugin
>>> Step 4: generating actions ...
>>> Step 4: interacting with world model ...
>>>>>>>>>>>>>>>>>>>>>>>>
>>> Step 5: generating actions ...
>>> Step 5: interacting with world model ...
>>>>>>>>>>>>>>>>>>>>>>>>
>>> Step 5: generating actions ...
>>> Step 5: interacting with world model ...
>>>>>>>>>>>>>>>>>>>>>>>>