Checkpoint step 20
Browse files- training_config.json +45 -0
training_config.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"learning_rate": 0.0001,
|
| 3 |
+
"weight_decay": 0.01,
|
| 4 |
+
"betas": [
|
| 5 |
+
0.9,
|
| 6 |
+
0.95
|
| 7 |
+
],
|
| 8 |
+
"grad_clip": 1.0,
|
| 9 |
+
"warmup_steps": 5,
|
| 10 |
+
"total_steps": 20,
|
| 11 |
+
"gradient_accumulation": 1,
|
| 12 |
+
"ema_decay": 0.9999,
|
| 13 |
+
"shift": 3.0,
|
| 14 |
+
"use_snr_weighting": true,
|
| 15 |
+
"snr_gamma": 5.0,
|
| 16 |
+
"use_huber_loss": false,
|
| 17 |
+
"huber_delta": 0.1,
|
| 18 |
+
"enable_lune": true,
|
| 19 |
+
"lune_weight": 0.1,
|
| 20 |
+
"lune_warmup_steps": 5,
|
| 21 |
+
"lune_dropout": 0.1,
|
| 22 |
+
"lune_mode": "cosine",
|
| 23 |
+
"enable_sol": true,
|
| 24 |
+
"sol_weight": 0.05,
|
| 25 |
+
"sol_warmup_steps": 10,
|
| 26 |
+
"use_spatial_weighting": false,
|
| 27 |
+
"text_dropout": 0.0,
|
| 28 |
+
"checkpoint_dir": "/tmp/tinyflux_e2e_9apre6n0/checkpoints",
|
| 29 |
+
"save_every_steps": 10,
|
| 30 |
+
"keep_last_n_steps": 2,
|
| 31 |
+
"save_every_epochs": 0,
|
| 32 |
+
"keep_last_n_epochs": 3,
|
| 33 |
+
"log_every": 5,
|
| 34 |
+
"tensorboard_dir": "/tmp/tinyflux_e2e_9apre6n0/tensorboard",
|
| 35 |
+
"sample_every": 10,
|
| 36 |
+
"sample_prompts": [
|
| 37 |
+
"a realistic cat, high quality",
|
| 38 |
+
"a watercolor dog"
|
| 39 |
+
],
|
| 40 |
+
"sample_dir": "/tmp/tinyflux_e2e_9apre6n0/samples",
|
| 41 |
+
"dtype": "torch.bfloat16",
|
| 42 |
+
"hf_repo_id": "AbstractPhil/tinyflux-test-e2e",
|
| 43 |
+
"upload_every_steps": 20,
|
| 44 |
+
"upload_every_epochs": 0
|
| 45 |
+
}
|