Upload z_image_base LoRA final checkpoint (epoch 80)

README.md

@@ -0,0 +1,18 @@
+---
+tags:
+- lora
+- diffusion-pipe
+- z_image_base
+base_model: z-image/z-image-base
+---
+
+# romanri85/zimage-base-lora
+
+LoRA trained with [diffusion-pipe](https://github.com/tdrussell/diffusion-pipe) via serverless RunPod worker.
+
+## Training Details
+
+- **Model type**: z_image_base
+- **Epochs**: 80
+- **LoRA rank**: 32
+- **Learning rate**: N/A

adapter_config.json

@@ -0,0 +1,74 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": null,
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "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",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "4.adaLN_modulation.0",
+    "15.adaLN_modulation.0",
+    "20.adaLN_modulation.0",
+    "16.adaLN_modulation.0",
+    "10.adaLN_modulation.0",
+    "26.adaLN_modulation.0",
+    "27.adaLN_modulation.0",
+    "18.adaLN_modulation.0",
+    "23.adaLN_modulation.0",
+    "29.adaLN_modulation.0",
+    "13.adaLN_modulation.0",
+    "14.adaLN_modulation.0",
+    "7.adaLN_modulation.0",
+    "w1",
+    "2.adaLN_modulation.0",
+    "22.adaLN_modulation.0",
+    "8.adaLN_modulation.0",
+    "qkv",
+    "17.adaLN_modulation.0",
+    "19.adaLN_modulation.0",
+    "w2",
+    "0.adaLN_modulation.0",
+    "out",
+    "28.adaLN_modulation.0",
+    "3.adaLN_modulation.0",
+    "25.adaLN_modulation.0",
+    "9.adaLN_modulation.0",
+    "11.adaLN_modulation.0",
+    "24.adaLN_modulation.0",
+    "5.adaLN_modulation.0",
+    "12.adaLN_modulation.0",
+    "21.adaLN_modulation.0",
+    "1.adaLN_modulation.0",
+    "w3",
+    "6.adaLN_modulation.0"
+  ],
+  "target_parameters": null,
+  "task_type": null,
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7ef37f2bc69573a8e265ea50a503ae1347ea41cc81fa37dd1750d0e6db6b50a
+size 174083128

training_config.toml

@@ -0,0 +1,29 @@
+output_dir = "/tmp/job_038749be-6705-42f5-99f9-a94f00a08cf4/training_outputs/z_image_base_lora"
+save_every_n_epochs = 10
+epochs = 80
+pipeline_stages = 1
+micro_batch_size_per_gpu = 1
+gradient_accumulation_steps = 1
+activation_checkpointing = true
+dataset = "/tmp/job_038749be-6705-42f5-99f9-a94f00a08cf4/dataset.toml"
+
+[model]
+type = "z_image"
+diffusion_model = "/workspace/models/models/z_image_base/z_image_bf16.safetensors"
+vae = "/workspace/models/models/z_image/ae.safetensors"
+dtype = "bfloat16"
+[[model.text_encoders]]
+path = "/workspace/models/models/z_image/qwen_3_4b.safetensors"
+type = "lumina2"
+
+[adapter]
+type = "lora"
+rank = 32
+dtype = "bfloat16"
+
+[optimizer]
+type = "adamw_optimi"
+lr = 0.0002
+betas = [ 0.9, 0.99,]
+weight_decay = 0.01
+eps = 1e-8

trigger_word.txt

@@ -0,0 +1 @@
+t3es9s5a1