File size: 10,556 Bytes
d0486d0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 | # Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES
#
# Licensed under the Apache License, Version 2.0 (the "License").
# You may not use this file except in compliance with the License.
# To view a copy of this license, visit http://www.apache.org/licenses/LICENSE-2.0
#
# No warranties are given. The work is provided "AS IS", without warranty of any kind, express or implied.
#
# SPDX-License-Identifier: Apache-2.0
import argparse
import os
from typing import List
import torch
import torch.distributed as dist
from omegaconf import OmegaConf
from tqdm import tqdm
from torch.utils.data import DataLoader, SequentialSampler
from torch.utils.data.distributed import DistributedSampler
from torchvision.io import write_video
from torchvision import transforms # noqa: F401
from einops import rearrange
from utils.misc import set_seed
from utils.distributed import barrier
from utils.memory import gpu, get_cuda_free_memory_gb, DynamicSwapInstaller
from pipeline.interactive_causal_inference import (
InteractiveCausalInferencePipeline,
)
from utils.dataset import MultiTextDataset
# ----------------------------- Argument parsing -----------------------------
parser = argparse.ArgumentParser("Interactive causal inference")
parser.add_argument("--config_path", type=str, help="Path to the config file")
args = parser.parse_args()
config = OmegaConf.load(args.config_path)
# ----------------------------- Distributed setup -----------------------------
if "LOCAL_RANK" in os.environ:
os.environ["NCCL_CROSS_NIC"] = "1"
os.environ["NCCL_DEBUG"] = os.environ.get("NCCL_DEBUG", "INFO")
os.environ["NCCL_TIMEOUT"] = os.environ.get("NCCL_TIMEOUT", "1800")
local_rank = int(os.environ["LOCAL_RANK"])
world_size = int(os.environ.get("WORLD_SIZE", "1"))
rank = int(os.environ.get("RANK", str(local_rank)))
# Set device first
torch.cuda.set_device(local_rank)
device = torch.device(f"cuda:{local_rank}")
# Initialize process group with backend and timeout
if not dist.is_initialized():
dist.init_process_group(
backend="nccl",
rank=rank,
world_size=world_size,
timeout=torch.distributed.constants.default_pg_timeout
)
set_seed(config.seed + local_rank)
print(f"[Rank {rank}] Initialized distributed processing on device {device}")
else:
local_rank = 0
rank = 0
device = torch.device("cuda")
set_seed(config.seed)
print(f"Single GPU mode on device {device}")
low_memory = get_cuda_free_memory_gb(device) < 40
torch.set_grad_enabled(False)
pipeline = InteractiveCausalInferencePipeline(config, device=device)
if config.generator_ckpt:
state_dict = torch.load(config.generator_ckpt, map_location="cpu")
raw_gen_state_dict = state_dict["generator_ema" if config.use_ema else "generator"]
if config.use_ema:
def _clean_key(name: str) -> str:
return name.replace("_fsdp_wrapped_module.", "")
cleaned_state_dict = {_clean_key(k): v for k, v in raw_gen_state_dict.items()}
missing, unexpected = pipeline.generator.load_state_dict(
cleaned_state_dict, strict=False
)
if local_rank == 0:
if missing:
print(f"[Warning] {len(missing)} parameters missing: {missing[:8]} ...")
if unexpected:
print(f"[Warning] {len(unexpected)} unexpected params: {unexpected[:8]} ...")
else:
pipeline.generator.load_state_dict(raw_gen_state_dict)
# --------------------------- LoRA support (optional) ---------------------------
from utils.lora_utils import configure_lora_for_model
import peft
pipeline.is_lora_enabled = False
if getattr(config, "adapter", None) and configure_lora_for_model is not None:
if local_rank == 0:
print(f"LoRA enabled with config: {config.adapter}")
print("Applying LoRA to generator (inference)...")
# After loading base weights, apply LoRA wrapper to the generator's transformer model
pipeline.generator.model = configure_lora_for_model(
pipeline.generator.model,
model_name="generator",
lora_config=config.adapter,
is_main_process=(local_rank == 0),
)
# Load LoRA weights (if lora_ckpt is provided)
lora_ckpt_path = getattr(config, "lora_ckpt", None)
if lora_ckpt_path:
if local_rank == 0:
print(f"Loading LoRA checkpoint from {lora_ckpt_path}")
lora_checkpoint = torch.load(lora_ckpt_path, map_location="cpu")
# Support both formats: containing the `generator_lora` key or a raw LoRA state dict
if isinstance(lora_checkpoint, dict) and "generator_lora" in lora_checkpoint:
peft.set_peft_model_state_dict(pipeline.generator.model, lora_checkpoint["generator_lora"]) # type: ignore
else:
peft.set_peft_model_state_dict(pipeline.generator.model, lora_checkpoint) # type: ignore
if local_rank == 0:
print("LoRA weights loaded for generator")
else:
if local_rank == 0:
print("No LoRA checkpoint specified; using base weights with LoRA adapters initialized")
pipeline.is_lora_enabled = True
# Move pipeline to appropriate dtype and device
print("dtype", pipeline.generator.model.dtype)
pipeline = pipeline.to(dtype=torch.bfloat16)
if low_memory:
DynamicSwapInstaller.install_model(pipeline.text_encoder, device=device)
pipeline.generator.to(device=device)
pipeline.vae.to(device=device)
# ----------------------------- Build dataset -----------------------------
# Parse switch_frame_indices
if isinstance(config.switch_frame_indices, int):
switch_frame_indices: List[int] = [int(config.switch_frame_indices)]
else:
switch_frame_indices: List[int] = [
int(x) for x in str(config.switch_frame_indices).split(",") if str(x).strip()
]
# Create dataset
dataset = MultiTextDataset(config.data_path)
# Validate number of segments & switch_frame_indices length
num_segments = len(dataset[0]["prompts_list"])
assert len(switch_frame_indices) == num_segments - 1, (
"The number of switch_frame_indices should be the number of prompt segments minus 1"
)
print("Number of segments:", num_segments)
print("Switch frame indices:", switch_frame_indices)
num_prompts_total = len(dataset)
print(f"Number of prompt lines: {num_prompts_total}")
if dist.is_initialized():
sampler = DistributedSampler(dataset, shuffle=False, drop_last=True)
else:
sampler = SequentialSampler(dataset)
dataloader = DataLoader(dataset, batch_size=1, sampler=sampler, num_workers=0, drop_last=False)
# Create output directory
if local_rank == 0:
os.makedirs(config.output_folder, exist_ok=True)
if dist.is_initialized():
dist.barrier()
# [episode run] episode_id for each JSONL line (same order as MultiTextDataset), timing/skip helpers
import json as _json, datetime as _dt, subprocess as _sp
with open(config.data_path, encoding="utf-8") as _f:
_episode_ids = [_json.loads(l)["episode_id"] for l in _f if l.strip()]
assert len(_episode_ids) == len(dataset)
def _ts():
return _dt.datetime.utcnow().strftime("%Y-%m-%dT%H:%M:%S.%fZ")
def _nframes(path):
r = _sp.run(["ffprobe", "-v", "error", "-count_frames", "-select_streams", "v:0", "-show_entries",
"stream=nb_read_frames", "-of", "csv=p=0", path], capture_output=True, text=True)
try:
return int(r.stdout.strip())
except ValueError:
return -1
_want_frames = 4 * config.num_output_frames - 3
# ----------------------------- Inference loop -----------------------------
for i, batch_data in tqdm(enumerate(dataloader), disable=(local_rank != 0)):
idx = batch_data["idx"].item()
prompts_list: List[str] = batch_data["prompts_list"] # type: ignore
sampled_noise = torch.randn(
[
config.num_samples,
config.num_output_frames,
16,
60,
104,
],
device=device,
dtype=torch.bfloat16,
)
# [episode run] skip finished outputs (noise is still drawn above, so the RNG stream matches an uninterrupted run)
episode_id = _episode_ids[idx]
_out = os.path.join(config.output_folder, f"{episode_id}.mp4")
if os.path.exists(_out) and _nframes(_out) == _want_frames:
print(f"{_ts()} [EPISODE] skip {episode_id} (exists, {_want_frames} frames)", flush=True)
continue
# [episode run] multi-GPU mode: claim the episode atomically (noise was already drawn, so outputs are identical)
if os.environ.get("LL_CLAIM_DIR"):
try:
os.mkdir(os.path.join(os.environ["LL_CLAIM_DIR"], episode_id))
except FileExistsError:
print(f"{_ts()} [EPISODE] skip {episode_id} (claimed by another worker)", flush=True)
continue
print(f"{_ts()} [EPISODE] start {episode_id} seed={config.seed} (set once per process)", flush=True)
torch.cuda.reset_peak_memory_stats()
video = pipeline.inference(
noise=sampled_noise,
text_prompts_list=prompts_list,
switch_frame_indices=switch_frame_indices,
return_latents=False,
)
current_video = rearrange(video, "b t c h w -> b t h w c").cpu() * 255.0
if dist.is_initialized():
rank = dist.get_rank()
else:
rank = 0
# Determine model type for filename
if hasattr(pipeline, 'is_lora_enabled') and pipeline.is_lora_enabled:
model_type = "lora"
elif getattr(config, 'use_ema', False):
model_type = "ema"
else:
model_type = "regular"
for seed_idx in range(config.num_samples):
if True: # [episode run] name output by episode_id
output_path = _out
elif config.save_with_index:
output_path = os.path.join(config.output_folder, f"rank{rank}-{idx}-{seed_idx}_{model_type}.mp4")
else:
# Use the first prompt segment as the filename prefix to avoid overly long names
short_name = prompts_list[0][0][:100].replace("/", "_")
output_path = os.path.join(config.output_folder, f"rank{rank}-{short_name}-{seed_idx}_{model_type}.mp4")
write_video(output_path, current_video[seed_idx].to(torch.uint8), fps=16)
print(f"{_ts()} [EPISODE] saved {episode_id} -> {output_path} frames={current_video.shape[1]} peak_mem_gb={torch.cuda.max_memory_allocated()/2**30:.1f}", flush=True)
if config.inference_iter != -1 and i >= config.inference_iter:
break
if dist.is_initialized():
dist.destroy_process_group() |