Create config.json

config.json

@@ -0,0 +1,64 @@
+{
+ "ngpus": 8,
+ "tokens": 50257,
+ "training": {
+  "batch_size": 512,
+  "accum": 1,
+  "n_iters": 1300001,
+  "snapshot_freq": 50000,
+  "log_freq": 50,
+  "eval_freq": 100,
+  "snapshot_freq_for_preemption": 10000,
+  "weight": "standard",
+  "snapshot_sampling": true,
+  "ema": 0.9999
+ },
+ "data": {
+  "train": "openwebtext",
+  "valid": "wikitext103",
+  "cache_dir": "data"
+ },
+ "graph": {
+  "type": "absorb",
+  "file": "data",
+  "report_all": false
+ },
+ "noise": {
+  "type": "loglinear",
+  "sigma_min": 0.0001,
+  "sigma_max": 20
+ },
+ "sampling": {
+  "predictor": "euler",
+  "steps": 128,
+  "noise_removal": true
+ },
+ "eval": {
+  "batch_size": 512,
+  "perplexity": true,
+  "perplexity_batch_size": 32
+ },
+ "optim": {
+  "weight_decay": 0,
+  "optimizer": "AdamW",
+  "lr": 0.0003,
+  "beta1": 0.9,
+  "beta2": 0.999,
+  "eps": 1e-08,
+  "warmup": 2500,
+  "grad_clip": 1.0
+ },
+ "model": {
+  "name": "small",
+  "type": "ddit",
+  "hidden_size": 768,
+  "cond_dim": 128,
+  "length": 1024,
+  "n_blocks": 12,
+  "n_heads": 12,
+  "scale_by_sigma": true,
+  "dropout": 0.1,
+  "score_strategy": 0
+ },
+ "work_dir": "absorb_small"
+}