Upload add_sub_baseline_50K
Browse files
add_sub_baseline_50K/config.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SorlModelWrapper"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": null,
|
| 8 |
+
"dtype": "float32",
|
| 9 |
+
"eos_token_id": null,
|
| 10 |
+
"head_dim": 128,
|
| 11 |
+
"hidden_act": "silu",
|
| 12 |
+
"hidden_size": 512,
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 2048,
|
| 15 |
+
"layer_types": [
|
| 16 |
+
"full_attention",
|
| 17 |
+
"full_attention",
|
| 18 |
+
"full_attention"
|
| 19 |
+
],
|
| 20 |
+
"max_position_embeddings": 128,
|
| 21 |
+
"max_window_layers": 28,
|
| 22 |
+
"model_type": "qwen3",
|
| 23 |
+
"num_attention_heads": 4,
|
| 24 |
+
"num_hidden_layers": 3,
|
| 25 |
+
"num_key_value_heads": 4,
|
| 26 |
+
"pad_token_id": null,
|
| 27 |
+
"rms_norm_eps": 1e-06,
|
| 28 |
+
"rope_parameters": {
|
| 29 |
+
"rope_theta": 10000.0,
|
| 30 |
+
"rope_type": "default"
|
| 31 |
+
},
|
| 32 |
+
"sliding_window": null,
|
| 33 |
+
"tie_word_embeddings": false,
|
| 34 |
+
"transformers_version": "5.5.0",
|
| 35 |
+
"use_cache": true,
|
| 36 |
+
"use_sliding_window": false,
|
| 37 |
+
"vocab_size": 151645
|
| 38 |
+
}
|
add_sub_baseline_50K/generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"output_attentions": false,
|
| 4 |
+
"output_hidden_states": false,
|
| 5 |
+
"transformers_version": "5.5.0",
|
| 6 |
+
"use_cache": true
|
| 7 |
+
}
|
add_sub_baseline_50K/metrics.json
ADDED
|
@@ -0,0 +1,197 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"history": {
|
| 3 |
+
"step": [
|
| 4 |
+
50,
|
| 5 |
+
100,
|
| 6 |
+
150,
|
| 7 |
+
200,
|
| 8 |
+
250,
|
| 9 |
+
300,
|
| 10 |
+
350,
|
| 11 |
+
400,
|
| 12 |
+
450,
|
| 13 |
+
500,
|
| 14 |
+
550,
|
| 15 |
+
600,
|
| 16 |
+
650,
|
| 17 |
+
700,
|
| 18 |
+
750,
|
| 19 |
+
800,
|
| 20 |
+
850,
|
| 21 |
+
900,
|
| 22 |
+
950,
|
| 23 |
+
1000,
|
| 24 |
+
1050,
|
| 25 |
+
1100,
|
| 26 |
+
1150,
|
| 27 |
+
1200,
|
| 28 |
+
1250,
|
| 29 |
+
1300,
|
| 30 |
+
1350,
|
| 31 |
+
1400,
|
| 32 |
+
1450,
|
| 33 |
+
1500,
|
| 34 |
+
1550,
|
| 35 |
+
1600,
|
| 36 |
+
1650,
|
| 37 |
+
1700,
|
| 38 |
+
1750,
|
| 39 |
+
1800,
|
| 40 |
+
1850,
|
| 41 |
+
1900,
|
| 42 |
+
1950,
|
| 43 |
+
2000,
|
| 44 |
+
2050,
|
| 45 |
+
2100,
|
| 46 |
+
2150,
|
| 47 |
+
2200,
|
| 48 |
+
2250,
|
| 49 |
+
2300
|
| 50 |
+
],
|
| 51 |
+
"loss": [
|
| 52 |
+
9.407320022583008,
|
| 53 |
+
7.224453926086426,
|
| 54 |
+
6.355410099029541,
|
| 55 |
+
5.141878604888916,
|
| 56 |
+
3.429792642593384,
|
| 57 |
+
2.3642735481262207,
|
| 58 |
+
1.9619863033294678,
|
| 59 |
+
1.8256255388259888,
|
| 60 |
+
1.7376317977905273,
|
| 61 |
+
1.3590067625045776,
|
| 62 |
+
0.5938534736633301,
|
| 63 |
+
0.4759748876094818,
|
| 64 |
+
0.3052070438861847,
|
| 65 |
+
0.1909075677394867,
|
| 66 |
+
0.17617182433605194,
|
| 67 |
+
0.1818520724773407,
|
| 68 |
+
0.11216778308153152,
|
| 69 |
+
0.12447719275951385,
|
| 70 |
+
0.07145003229379654,
|
| 71 |
+
0.08844210207462311,
|
| 72 |
+
0.06295949220657349,
|
| 73 |
+
0.03105434775352478,
|
| 74 |
+
0.038688015192747116,
|
| 75 |
+
0.04730251058936119,
|
| 76 |
+
0.04191047325730324,
|
| 77 |
+
0.05501880869269371,
|
| 78 |
+
0.0545913390815258,
|
| 79 |
+
0.03572807461023331,
|
| 80 |
+
0.022829385474324226,
|
| 81 |
+
0.0187111496925354,
|
| 82 |
+
0.025716397911310196,
|
| 83 |
+
0.0280207097530365,
|
| 84 |
+
0.018566560000181198,
|
| 85 |
+
0.01017980556935072,
|
| 86 |
+
0.021484609693288803,
|
| 87 |
+
0.006599620915949345,
|
| 88 |
+
0.005485634785145521,
|
| 89 |
+
0.012738294899463654,
|
| 90 |
+
0.004766442347317934,
|
| 91 |
+
0.006911102682352066,
|
| 92 |
+
0.007414546329528093,
|
| 93 |
+
0.011577660217881203,
|
| 94 |
+
0.005165865179151297,
|
| 95 |
+
0.01003621332347393,
|
| 96 |
+
0.006265468895435333,
|
| 97 |
+
0.00291456445120275
|
| 98 |
+
],
|
| 99 |
+
"base_loss": [
|
| 100 |
+
9.407320022583008,
|
| 101 |
+
7.224453926086426,
|
| 102 |
+
6.355410099029541,
|
| 103 |
+
5.141878604888916,
|
| 104 |
+
3.429792642593384,
|
| 105 |
+
2.3642735481262207,
|
| 106 |
+
1.9619863033294678,
|
| 107 |
+
1.8256255388259888,
|
| 108 |
+
1.7376317977905273,
|
| 109 |
+
1.3590067625045776,
|
| 110 |
+
0.5938534736633301,
|
| 111 |
+
0.4759748876094818,
|
| 112 |
+
0.3052070438861847,
|
| 113 |
+
0.1909075677394867,
|
| 114 |
+
0.17617182433605194,
|
| 115 |
+
0.1818520724773407,
|
| 116 |
+
0.11216778308153152,
|
| 117 |
+
0.12447719275951385,
|
| 118 |
+
0.07145003229379654,
|
| 119 |
+
0.08844210207462311,
|
| 120 |
+
0.06295949220657349,
|
| 121 |
+
0.03105434775352478,
|
| 122 |
+
0.038688015192747116,
|
| 123 |
+
0.04730251058936119,
|
| 124 |
+
0.04191047325730324,
|
| 125 |
+
0.05501880869269371,
|
| 126 |
+
0.0545913390815258,
|
| 127 |
+
0.03572807461023331,
|
| 128 |
+
0.022829385474324226,
|
| 129 |
+
0.0187111496925354,
|
| 130 |
+
0.025716397911310196,
|
| 131 |
+
0.0280207097530365,
|
| 132 |
+
0.018566560000181198,
|
| 133 |
+
0.01017980556935072,
|
| 134 |
+
0.021484609693288803,
|
| 135 |
+
0.006599620915949345,
|
| 136 |
+
0.005485634785145521,
|
| 137 |
+
0.012738294899463654,
|
| 138 |
+
0.004766442347317934,
|
| 139 |
+
0.006911102682352066,
|
| 140 |
+
0.007414546329528093,
|
| 141 |
+
0.011577660217881203,
|
| 142 |
+
0.005165865179151297,
|
| 143 |
+
0.01003621332347393,
|
| 144 |
+
0.006265468895435333,
|
| 145 |
+
0.00291456445120275
|
| 146 |
+
],
|
| 147 |
+
"lr": [
|
| 148 |
+
8.358208955223882e-06,
|
| 149 |
+
1.6886993603411513e-05,
|
| 150 |
+
2.541577825159915e-05,
|
| 151 |
+
3.394456289978678e-05,
|
| 152 |
+
4.247334754797441e-05,
|
| 153 |
+
5.100213219616205e-05,
|
| 154 |
+
5.953091684434969e-05,
|
| 155 |
+
6.805970149253732e-05,
|
| 156 |
+
7.658848614072494e-05,
|
| 157 |
+
7.994958584913144e-05,
|
| 158 |
+
7.96419594562729e-05,
|
| 159 |
+
7.905686510146095e-05,
|
| 160 |
+
7.819839806506031e-05,
|
| 161 |
+
7.707256705860444e-05,
|
| 162 |
+
7.568725216771852e-05,
|
| 163 |
+
7.405214969663954e-05,
|
| 164 |
+
7.217870430038601e-05,
|
| 165 |
+
7.008002887960812e-05,
|
| 166 |
+
6.777081279880229e-05,
|
| 167 |
+
6.526721907030291e-05,
|
| 168 |
+
6.258677122369667e-05,
|
| 169 |
+
5.9748230652499966e-05,
|
| 170 |
+
5.677146529659303e-05,
|
| 171 |
+
5.3677310579548395e-05,
|
| 172 |
+
5.04874235742021e-05,
|
| 173 |
+
4.722413141721374e-05,
|
| 174 |
+
4.391027503361521e-05,
|
| 175 |
+
4.056904926517458e-05,
|
| 176 |
+
3.722384052157265e-05,
|
| 177 |
+
3.3898063090728774e-05,
|
| 178 |
+
3.061499525399694e-05,
|
| 179 |
+
2.7397616353319584e-05,
|
| 180 |
+
2.4268445950762915e-05,
|
| 181 |
+
2.1249386206212863e-05,
|
| 182 |
+
1.8361568576485003e-05,
|
| 183 |
+
1.5625205908855437e-05,
|
| 184 |
+
1.3059450964261466e-05,
|
| 185 |
+
1.0682262360418084e-05,
|
| 186 |
+
8.510278873161129e-06,
|
| 187 |
+
6.558702975826241e-06,
|
| 188 |
+
4.841194431812048e-06,
|
| 189 |
+
3.3697746851107672e-06,
|
| 190 |
+
2.1547427180102297e-06,
|
| 191 |
+
1.2046029649088297e-06,
|
| 192 |
+
5.260057867998169e-07,
|
| 193 |
+
1.2370092306488446e-07
|
| 194 |
+
]
|
| 195 |
+
},
|
| 196 |
+
"final_accuracy": 0.98
|
| 197 |
+
}
|
add_sub_baseline_50K/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5490f49379b81c1cae6704d8090541cdb0c1699eeb937ba75a009f520c0e01f8
|
| 3 |
+
size 671794850
|
add_sub_baseline_50K/train_config.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"mode": "baseline",
|
| 3 |
+
"ops": "add_sub",
|
| 4 |
+
"n_digits": 6,
|
| 5 |
+
"n_layer": 3,
|
| 6 |
+
"n_head": 4,
|
| 7 |
+
"n_embd": 512,
|
| 8 |
+
"abs_vocab": 0,
|
| 9 |
+
"K": 4,
|
| 10 |
+
"batch_size": 64,
|
| 11 |
+
"num_epochs": 3,
|
| 12 |
+
"dataset_size": 50000,
|
| 13 |
+
"lr": 8e-05,
|
| 14 |
+
"output_dir": "ckpt/r/add_sub_baseline_50K",
|
| 15 |
+
"device": "cuda",
|
| 16 |
+
"push_to_hub": true,
|
| 17 |
+
"no_wandb": false,
|
| 18 |
+
"n_params": 167871744,
|
| 19 |
+
"run_name": "add_sub_baseline_50K",
|
| 20 |
+
"git_commit": "9e4530548a98f8c7f5c14930ac4aec4886bb4b1b",
|
| 21 |
+
"timestamp": "2026-04-07T05:27:33.004161",
|
| 22 |
+
"tokenizer": "Qwen/Qwen3-0.6B",
|
| 23 |
+
"dataset_repo": "thoughtworks/arithmetic-sorl-data",
|
| 24 |
+
"dataset_config": "add_sub_6digit",
|
| 25 |
+
"model_repo": "thoughtworks/arithmetic-sorl",
|
| 26 |
+
"trainer_version": "sft",
|
| 27 |
+
"final_accuracy": 0.98
|
| 28 |
+
}
|