Upload gpt-gauss-ts-d1/meta_000001.json with huggingface_hub
Browse files
gpt-gauss-ts-d1/meta_000001.json
ADDED
|
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"step": 1,
|
| 3 |
+
"val_bpb": NaN,
|
| 4 |
+
"model_config": {
|
| 5 |
+
"sequence_len": 2048,
|
| 6 |
+
"vocab_size": 265,
|
| 7 |
+
"n_layer": 1,
|
| 8 |
+
"n_head": 16,
|
| 9 |
+
"n_kv_head": 16,
|
| 10 |
+
"n_embd": 128,
|
| 11 |
+
"adaptive_decay": false,
|
| 12 |
+
"rope": true,
|
| 13 |
+
"repetition_count": 1,
|
| 14 |
+
"mamba_layers": [],
|
| 15 |
+
"recurrent_vocab_layers": [],
|
| 16 |
+
"gauss_layers": [
|
| 17 |
+
0
|
| 18 |
+
],
|
| 19 |
+
"attention_time_decay": false,
|
| 20 |
+
"rva_viz": false
|
| 21 |
+
},
|
| 22 |
+
"user_config": {
|
| 23 |
+
"run": "dummy",
|
| 24 |
+
"device_type": "",
|
| 25 |
+
"depth": 1,
|
| 26 |
+
"repetition_count": 1,
|
| 27 |
+
"recurrent_vocab_layers": [],
|
| 28 |
+
"mamba_layers": [],
|
| 29 |
+
"gauss_layers": [
|
| 30 |
+
0
|
| 31 |
+
],
|
| 32 |
+
"attention_time_decay": false,
|
| 33 |
+
"adaptive_decay": false,
|
| 34 |
+
"rope": true,
|
| 35 |
+
"init_from_tag": "",
|
| 36 |
+
"rva_viz": false,
|
| 37 |
+
"resume_from_tag": "",
|
| 38 |
+
"resume_from_step": -1,
|
| 39 |
+
"resume_from_hub": true,
|
| 40 |
+
"resume_hub_repo_id": "ASzecsenyi/VQLM",
|
| 41 |
+
"resume_hub_subdir": "",
|
| 42 |
+
"resume_hub_repo_type": "model",
|
| 43 |
+
"num_iterations": -1,
|
| 44 |
+
"target_flops": -1.0,
|
| 45 |
+
"target_param_data_ratio": 1,
|
| 46 |
+
"num_epochs": -1,
|
| 47 |
+
"data": "tinystories_data",
|
| 48 |
+
"vocab_size": 265,
|
| 49 |
+
"max_seq_len": 2048,
|
| 50 |
+
"device_batch_size": 4,
|
| 51 |
+
"total_batch_size": -1,
|
| 52 |
+
"embedding_lr": 0.2,
|
| 53 |
+
"unembedding_lr": 0.004,
|
| 54 |
+
"weight_decay": 0.0,
|
| 55 |
+
"matrix_lr": 0.02,
|
| 56 |
+
"base_lr": 1.0,
|
| 57 |
+
"grad_clip": 1.0,
|
| 58 |
+
"warmup_ratio": 0.05,
|
| 59 |
+
"warmdown_ratio": 0.2,
|
| 60 |
+
"final_lr_frac": 0.0,
|
| 61 |
+
"eval_every": 2500,
|
| 62 |
+
"eval_tokens": 10485760,
|
| 63 |
+
"core_metric_every": 2000,
|
| 64 |
+
"core_metric_max_per_task": 500,
|
| 65 |
+
"sample_every": 2500,
|
| 66 |
+
"checkpoint_every": -1,
|
| 67 |
+
"max_checkpoints": 3,
|
| 68 |
+
"push_checkpoints_to_hub": true,
|
| 69 |
+
"hf_repo_id": "ASzecsenyi/VQLM",
|
| 70 |
+
"hf_repo_type": "model",
|
| 71 |
+
"model_tag": ""
|
| 72 |
+
},
|
| 73 |
+
"device_batch_size": 4,
|
| 74 |
+
"max_seq_len": 2048,
|
| 75 |
+
"num_iterations": 16,
|
| 76 |
+
"warmdown_ratio": 0.2,
|
| 77 |
+
"max_checkpoints": 3,
|
| 78 |
+
"push_checkpoints_to_hub": true,
|
| 79 |
+
"hf_repo_id": "ASzecsenyi/VQLM",
|
| 80 |
+
"hf_repo_type": "model"
|
| 81 |
+
}
|