Upload rva_ts_1x256_sample_dummy/meta_000035.json with huggingface_hub
Browse files
rva_ts_1x256_sample_dummy/meta_000035.json
CHANGED
|
@@ -43,7 +43,7 @@
|
|
| 43 |
"warmup_ratio": 0.0,
|
| 44 |
"warmdown_ratio": 0.4,
|
| 45 |
"final_lr_frac": 0.0,
|
| 46 |
-
"resume_from_step":
|
| 47 |
"eval_every": 250,
|
| 48 |
"eval_tokens": 10485760,
|
| 49 |
"core_metric_every": -1,
|
|
@@ -103,7 +103,7 @@
|
|
| 103 |
"warmup_ratio": 0.0,
|
| 104 |
"warmdown_ratio": 0.4,
|
| 105 |
"final_lr_frac": 0.0,
|
| 106 |
-
"resume_from_step":
|
| 107 |
"eval_every": 250,
|
| 108 |
"eval_tokens": 10485760,
|
| 109 |
"core_metric_every": -1,
|
|
@@ -121,7 +121,7 @@
|
|
| 121 |
"sequence_len": 1024,
|
| 122 |
"dataloader_state_dict": {
|
| 123 |
"pq_idx": 0,
|
| 124 |
-
"rg_idx":
|
| 125 |
},
|
| 126 |
"loop_state": {
|
| 127 |
"min_val_bpb": 3.310937841614675,
|
|
|
|
| 43 |
"warmup_ratio": 0.0,
|
| 44 |
"warmdown_ratio": 0.4,
|
| 45 |
"final_lr_frac": 0.0,
|
| 46 |
+
"resume_from_step": 35,
|
| 47 |
"eval_every": 250,
|
| 48 |
"eval_tokens": 10485760,
|
| 49 |
"core_metric_every": -1,
|
|
|
|
| 103 |
"warmup_ratio": 0.0,
|
| 104 |
"warmdown_ratio": 0.4,
|
| 105 |
"final_lr_frac": 0.0,
|
| 106 |
+
"resume_from_step": 35,
|
| 107 |
"eval_every": 250,
|
| 108 |
"eval_tokens": 10485760,
|
| 109 |
"core_metric_every": -1,
|
|
|
|
| 121 |
"sequence_len": 1024,
|
| 122 |
"dataloader_state_dict": {
|
| 123 |
"pq_idx": 0,
|
| 124 |
+
"rg_idx": 24
|
| 125 |
},
|
| 126 |
"loop_state": {
|
| 127 |
"min_val_bpb": 3.310937841614675,
|