Added PEFT and diffusion-pipe config data

adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": null,
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q",
+    "k",
+    "ffn.2",
+    "o",
+    "v",
+    "ffn.0"
+  ],
+  "task_type": null,
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

config.toml

@@ -0,0 +1,60 @@
+# Dataset config file.
+output_dir = '/root/outputs'
+dataset = 'configs/dataset.toml'
+
+# Training settings
+epochs = 50
+micro_batch_size_per_gpu = 5
+pipeline_stages = 1
+gradient_accumulation_steps = 1
+gradient_clipping = 1.0
+warmup_steps = 20
+
+# eval settings
+eval_every_n_epochs = 1
+eval_before_first_step = false
+eval_micro_batch_size_per_gpu = 5
+eval_gradient_accumulation_steps = 1
+
+# misc settings
+save_every_n_epochs = 2
+checkpoint_every_n_epochs = 2
+activation_checkpointing = true
+partition_method = 'parameters'
+save_dtype = 'bfloat16'
+caching_batch_size = 3
+steps_per_print = 1
+compile = true
+video_clip_mode = 'single_beginning'
+#blocks_to_swap = 15
+
+[model]
+type = 'wan'
+ckpt_path = '/root/outputs/models/Wan2.1-T2V-14B'
+#diffusers_path = '/root/outputs/models/FLUX.1-dev'
+#transformer_path = '/root/outputs/models/chroma/Chroma.safetensors'
+llm_path = '/root/outputs/models/Wan2.1-T2V-14B/models_t5_umt5-xxl-enc-bf16.pth'
+dtype = 'bfloat16'
+# You can optionally load the transformer in fp8 when training LoRAs.
+transformer_dtype = 'float8'
+timestep_sample_method = 'logit_normal'
+#flux_shift = true
+
+[adapter]
+type = 'lora'
+rank = 32
+dtype = 'bfloat16'
+#init_from_existing = '/root/outputs/5c6d31124f144544913effcc0a17e0ea/epoch10'
+
+[optimizer]
+type = 'adamw_optimi'
+lr = 2e-4
+betas = [0.9, 0.99]
+weight_decay = 0.01
+eps = 1e-8
+
+[monitoring]
+enable_wandb = true
+wandb_api_key = '316edc68e8c9d21674a9bb887bc5c599e2c94e35'
+wandb_tracker_name = 'wan'
+wandb_run_name = 'wan-kwis-v1'

wandb-metadata.json

@@ -0,0 +1,46 @@
+{
+  "os":  "Linux-4.4.0-x86_64-with-glibc2.36",
+  "python":  "CPython 3.12.7",
+  "startedAt":  "2025-08-07T17:50:41.273316Z",
+  "args":  [
+    "--local_rank=0",
+    "--deepspeed",
+    "--config",
+    "configs/config.toml"
+  ],
+  "program":  "/root/diffusion-pipe/train.py",
+  "codePath":  "train.py",
+  "codePathLocal":  "train.py",
+  "git":  {
+    "remote":  "https://github.com/tdrussell/diffusion-pipe.git",
+    "commit":  "6b65063a5085f799cc91508c735ce4cb6fc6b036"
+  },
+  "email":  "[email protected]",
+  "root":  "/root/outputs/models",
+  "host":  "modal",
+  "executable":  "/opt/conda/bin/python3.12",
+  "cpu_count":  17,
+  "cpu_count_logical":  17,
+  "gpu":  "NVIDIA L40S",
+  "gpu_count":  1,
+  "disk":  {
+    "/":  {
+      "total":  "549755813888",
+      "used":  "13826981888"
+    }
+  },
+  "memory":  {
+    "total":  "359713656832"
+  },
+  "gpu_nvidia":  [
+    {
+      "name":  "NVIDIA L40S",
+      "memoryTotal":  "48305799168",
+      "cudaCores":  18176,
+      "architecture":  "Ada",
+      "uuid":  "GPU-fa9be711-9328-649f-0757-e99aedf64bb8"
+    }
+  ],
+  "cudaVersion":  "12.9",
+  "writerId":  "54v4o419o3hllzfl28qql5njcy5i57ok"
+}