fabikru commited on
Commit
e0dca70
·
verified ·
1 Parent(s): 69805a0

new_fabikru/half-of-chembl-2025-randomized-smiles-cleaned_modern_smiles_bert_4.5M_character-level_MLM0.3

Browse files
Files changed (4) hide show
  1. README.md +43 -30
  2. config.json +4 -4
  3. model.safetensors +2 -2
  4. training_args.bin +1 -1
README.md CHANGED
@@ -16,8 +16,8 @@ should probably proofread and complete it, then remove this comment. -->
16
 
17
  This model is a fine-tuned version of [](https://huggingface.co/) on an unknown dataset.
18
  It achieves the following results on the evaluation set:
19
- - Loss: 0.1410
20
- - Accuracy: 0.9492
21
 
22
  ## Model description
23
 
@@ -36,7 +36,7 @@ More information needed
36
  ### Training hyperparameters
37
 
38
  The following hyperparameters were used during training:
39
- - learning_rate: 0.000834
40
  - train_batch_size: 256
41
  - eval_batch_size: 256
42
  - seed: 42
@@ -50,33 +50,46 @@ The following hyperparameters were used during training:
50
 
51
  | Training Loss | Epoch | Step | Validation Loss | Accuracy |
52
  |:-------------:|:-------:|:-----:|:---------------:|:--------:|
53
- | No log | 0 | 0 | 4.4485 | 0.0010 |
54
- | 0.415 | 0.4302 | 1953 | 0.3299 | 0.8849 |
55
- | 0.2874 | 0.8604 | 3906 | 0.2612 | 0.9074 |
56
- | 0.2486 | 1.2905 | 5859 | 0.2317 | 0.9179 |
57
- | 0.2308 | 1.7207 | 7812 | 0.2167 | 0.9231 |
58
- | 0.2161 | 2.1509 | 9765 | 0.2036 | 0.9273 |
59
- | 0.2057 | 2.5811 | 11718 | 0.1952 | 0.9303 |
60
- | 0.1976 | 3.0112 | 13671 | 0.1875 | 0.9331 |
61
- | 0.1908 | 3.4414 | 15624 | 0.1834 | 0.9344 |
62
- | 0.1854 | 3.8716 | 17577 | 0.1768 | 0.9365 |
63
- | 0.1799 | 4.3018 | 19530 | 0.1722 | 0.9386 |
64
- | 0.1776 | 4.7319 | 21483 | 0.1692 | 0.9392 |
65
- | 0.1728 | 5.1621 | 23436 | 0.1668 | 0.9400 |
66
- | 0.1698 | 5.5923 | 25389 | 0.1632 | 0.9416 |
67
- | 0.1672 | 6.0225 | 27342 | 0.1616 | 0.9418 |
68
- | 0.1648 | 6.4526 | 29295 | 0.1587 | 0.9430 |
69
- | 0.1616 | 6.8828 | 31248 | 0.1561 | 0.9439 |
70
- | 0.1591 | 7.3130 | 33201 | 0.1552 | 0.9438 |
71
- | 0.1585 | 7.7432 | 35154 | 0.1528 | 0.9450 |
72
- | 0.1546 | 8.1733 | 37107 | 0.1502 | 0.9461 |
73
- | 0.1522 | 8.6035 | 39060 | 0.1491 | 0.9466 |
74
- | 0.1521 | 9.0337 | 41013 | 0.1484 | 0.9468 |
75
- | 0.1501 | 9.4639 | 42966 | 0.1452 | 0.9478 |
76
- | 0.1472 | 9.8941 | 44919 | 0.1440 | 0.9481 |
77
- | 0.1455 | 10.3242 | 46872 | 0.1412 | 0.9494 |
78
- | 0.1443 | 10.7544 | 48825 | 0.1428 | 0.9486 |
79
- | 0.1433 | 11.1846 | 50778 | 0.1402 | 0.9494 |
 
 
 
 
 
 
 
 
 
 
 
 
 
80
 
81
 
82
  ### Framework versions
 
16
 
17
  This model is a fine-tuned version of [](https://huggingface.co/) on an unknown dataset.
18
  It achieves the following results on the evaluation set:
19
+ - Loss: 0.1940
20
+ - Accuracy: 0.9309
21
 
22
  ## Model description
23
 
 
36
  ### Training hyperparameters
37
 
38
  The following hyperparameters were used during training:
39
+ - learning_rate: 0.008141
40
  - train_batch_size: 256
41
  - eval_batch_size: 256
42
  - seed: 42
 
50
 
51
  | Training Loss | Epoch | Step | Validation Loss | Accuracy |
52
  |:-------------:|:-------:|:-----:|:---------------:|:--------:|
53
+ | No log | 0 | 0 | 4.3643 | 0.0040 |
54
+ | 0.5212 | 0.4302 | 1953 | 0.4259 | 0.8526 |
55
+ | 0.3855 | 0.8604 | 3906 | 0.3514 | 0.8772 |
56
+ | 0.3391 | 1.2905 | 5859 | 0.3167 | 0.8893 |
57
+ | 0.3159 | 1.7207 | 7812 | 0.2958 | 0.8961 |
58
+ | 0.2969 | 2.1509 | 9765 | 0.2806 | 0.9011 |
59
+ | 0.2842 | 2.5811 | 11718 | 0.2707 | 0.9045 |
60
+ | 0.2741 | 3.0112 | 13671 | 0.2610 | 0.9077 |
61
+ | 0.2652 | 3.4414 | 15624 | 0.2548 | 0.9097 |
62
+ | 0.2596 | 3.8716 | 17577 | 0.2476 | 0.9123 |
63
+ | 0.253 | 4.3018 | 19530 | 0.2441 | 0.9135 |
64
+ | 0.2511 | 4.7319 | 21483 | 0.2398 | 0.9145 |
65
+ | 0.2456 | 5.1621 | 23436 | 0.2364 | 0.9158 |
66
+ | 0.2415 | 5.5923 | 25389 | 0.2319 | 0.9173 |
67
+ | 0.2389 | 6.0225 | 27342 | 0.2290 | 0.9185 |
68
+ | 0.236 | 6.4526 | 29295 | 0.2274 | 0.9192 |
69
+ | 0.2323 | 6.8828 | 31248 | 0.2240 | 0.9202 |
70
+ | 0.2296 | 7.3130 | 33201 | 0.2226 | 0.9206 |
71
+ | 0.2282 | 7.7432 | 35154 | 0.2200 | 0.9216 |
72
+ | 0.2254 | 8.1733 | 37107 | 0.2165 | 0.9228 |
73
+ | 0.2221 | 8.6035 | 39060 | 0.2157 | 0.9230 |
74
+ | 0.2216 | 9.0337 | 41013 | 0.2144 | 0.9235 |
75
+ | 0.2204 | 9.4639 | 42966 | 0.2126 | 0.9243 |
76
+ | 0.2172 | 9.8941 | 44919 | 0.2103 | 0.9249 |
77
+ | 0.2149 | 10.3242 | 46872 | 0.2084 | 0.9254 |
78
+ | 0.2142 | 10.7544 | 48825 | 0.2091 | 0.9254 |
79
+ | 0.213 | 11.1846 | 50778 | 0.2065 | 0.9261 |
80
+ | 0.2117 | 11.6148 | 52731 | 0.2054 | 0.9267 |
81
+ | 0.2118 | 12.0449 | 54684 | 0.2035 | 0.9274 |
82
+ | 0.2095 | 12.4751 | 56637 | 0.2042 | 0.9272 |
83
+ | 0.2083 | 12.9053 | 58590 | 0.2024 | 0.9277 |
84
+ | 0.2069 | 13.3355 | 60543 | 0.2027 | 0.9278 |
85
+ | 0.2071 | 13.7656 | 62496 | 0.2004 | 0.9283 |
86
+ | 0.2063 | 14.1958 | 64449 | 0.1991 | 0.9289 |
87
+ | 0.2043 | 14.6260 | 66402 | 0.1998 | 0.9284 |
88
+ | 0.2054 | 15.0562 | 68355 | 0.1972 | 0.9295 |
89
+ | 0.2026 | 15.4863 | 70308 | 0.1973 | 0.9294 |
90
+ | 0.2024 | 15.9165 | 72261 | 0.1956 | 0.9301 |
91
+ | 0.2013 | 16.3467 | 74214 | 0.1952 | 0.9304 |
92
+ | 0.2007 | 16.7769 | 76167 | 0.1945 | 0.9305 |
93
 
94
 
95
  ### Framework versions
config.json CHANGED
@@ -17,10 +17,10 @@
17
  "global_attn_every_n_layers": 1,
18
  "global_rope_theta": 160000.0,
19
  "hidden_activation": "gelu",
20
- "hidden_size": 768,
21
  "initializer_cutoff_factor": 2.0,
22
  "initializer_range": 0.02,
23
- "intermediate_size": 1152,
24
  "local_attention": 128,
25
  "local_rope_theta": 10000.0,
26
  "max_position_embeddings": 502,
@@ -29,8 +29,8 @@
29
  "model_type": "modernbert",
30
  "norm_bias": false,
31
  "norm_eps": 1e-05,
32
- "num_attention_heads": 12,
33
- "num_hidden_layers": 22,
34
  "pad_token_id": 1,
35
  "reference_compile": true,
36
  "repad_logits_with_grad": false,
 
17
  "global_attn_every_n_layers": 1,
18
  "global_rope_theta": 160000.0,
19
  "hidden_activation": "gelu",
20
+ "hidden_size": 256,
21
  "initializer_cutoff_factor": 2.0,
22
  "initializer_range": 0.02,
23
+ "intermediate_size": 384,
24
  "local_attention": 128,
25
  "local_rope_theta": 10000.0,
26
  "max_position_embeddings": 502,
 
29
  "model_type": "modernbert",
30
  "norm_bias": false,
31
  "norm_eps": 1e-05,
32
+ "num_attention_heads": 4,
33
+ "num_hidden_layers": 8,
34
  "pad_token_id": 1,
35
  "reference_compile": true,
36
  "repad_logits_with_grad": false,
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dd6320094e406e36b2ec9f34edd800c714aea3e4282166982ff36a93824840f2
3
- size 443955224
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:788ba3089867da43cd59920d146c37ea24e323d7d71dceafed99a689abf07c7c
3
+ size 18195880
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f92ffe3fcc8f350e3d943880832ccfd60e97e58375a433123b97b1c948cc28e6
3
  size 5905
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44106c6ba5dc86ae6f86209e5f3e3e6c5e4acce8d60d4a87d0cbc225634500a1
3
  size 5905