Hugging Face's logo

skfrost19
/
reranker-ModernBERT-base-msmarco-bce-Lion.Cosine-ep-1-3

Text Ranking
sentence-transformers
Safetensors
English
modernbert
cross-encoder
Generated from Trainer
dataset_size:1990000
loss:BinaryCrossEntropyLoss
Eval Results
Model card Files Community
skfrost19 commited on
Commit
83f8efd
·
verified ·
1 Parent(s): b79cf38

Update model with fully trained weights from checkpoint-31092

Browse files
Files changed (4) hide show
  1. README.md +33 -35
  2. config.json +2 -3
  3. model.safetensors +1 -1
  4. tokenizer_config.json +7 -0
README.md CHANGED
@@ -27,13 +27,13 @@ model-index:
27
  type: NanoMSMARCO_R100
28
  metrics:
29
  - type: map
30
- value: 0.6322
31
  name: Map
32
  - type: mrr@10
33
- value: 0.6257
34
  name: Mrr@10
35
  - type: ndcg@10
36
- value: 0.6902
37
  name: Ndcg@10
38
  - task:
39
  type: cross-encoder-reranking
@@ -43,13 +43,13 @@ model-index:
43
  type: NanoNFCorpus_R100
44
  metrics:
45
  - type: map
46
- value: 0.3433
47
  name: Map
48
  - type: mrr@10
49
- value: 0.5436
50
  name: Mrr@10
51
  - type: ndcg@10
52
- value: 0.3796
53
  name: Ndcg@10
54
  - task:
55
  type: cross-encoder-reranking
@@ -59,13 +59,13 @@ model-index:
59
  type: NanoNQ_R100
60
  metrics:
61
  - type: map
62
- value: 0.6859
63
  name: Map
64
  - type: mrr@10
65
- value: 0.7007
66
  name: Mrr@10
67
  - type: ndcg@10
68
- value: 0.7323
69
  name: Ndcg@10
70
  - task:
71
  type: cross-encoder-nano-beir
@@ -75,13 +75,13 @@ model-index:
75
  type: NanoBEIR_R100_mean
76
  metrics:
77
  - type: map
78
- value: 0.5538
79
  name: Map
80
  - type: mrr@10
81
- value: 0.6233
82
  name: Mrr@10
83
  - type: ndcg@10
84
- value: 0.6007
85
  name: Ndcg@10
86
  ---
87
 
@@ -191,9 +191,9 @@ You can finetune this model on your own dataset.
191
 
192
  | Metric | NanoMSMARCO_R100 | NanoNFCorpus_R100 | NanoNQ_R100 |
193
  |:------------|:---------------------|:---------------------|:---------------------|
194
- | map | 0.6322 (+0.1426) | 0.3433 (+0.0823) | 0.6859 (+0.2663) |
195
- | mrr@10 | 0.6257 (+0.1482) | 0.5436 (+0.0437) | 0.7007 (+0.2740) |
196
- | **ndcg@10** | **0.6902 (+0.1498)** | **0.3796 (+0.0546)** | **0.7323 (+0.2316)** |
197
 
198
  #### Cross Encoder Nano BEIR
199
 
@@ -214,9 +214,9 @@ You can finetune this model on your own dataset.
214
 
215
  | Metric | Value |
216
  |:------------|:---------------------|
217
- | map | 0.5538 (+0.1637) |
218
- | mrr@10 | 0.6233 (+0.1553) |
219
- | **ndcg@10** | **0.6007 (+0.1453)** |
220
 
221
  <!--
222
  ## Bias, Risks and Limitations
@@ -421,23 +421,21 @@ You can finetune this model on your own dataset.
421
  </details>
422
 
423
  ### Training Logs
424
- | Epoch | Step | Training Loss | Validation Loss | NanoMSMARCO_R100_ndcg@10 | NanoNFCorpus_R100_ndcg@10 | NanoNQ_R100_ndcg@10 | NanoBEIR_R100_mean_ndcg@10 |
425
- |:-------:|:---------:|:-------------:|:---------------:|:------------------------:|:-------------------------:|:--------------------:|:--------------------------:|
426
- | -1 | -1 | - | - | 0.0186 (-0.5218) | 0.2940 (-0.0310) | 0.0428 (-0.4579) | 0.1185 (-0.3369) |
427
- | 0.0001 | 1 | 0.7403 | - | - | - | - | - |
428
- | 0.3860 | 4000 | 0.2255 | - | - | - | - | - |
429
- | 0.7719 | 8000 | 0.153 | - | - | - | - | - |
430
- | **1.0** | **10364** | **-** | **0.1429** | **0.6902 (+0.1498)** | **0.3796 (+0.0546)** | **0.7323 (+0.2316)** | **0.6007 (+0.1453)** |
431
- | 1.1579 | 12000 | 0.1356 | - | - | - | - | - |
432
- | 1.5438 | 16000 | 0.122 | - | - | - | - | - |
433
- | 1.9298 | 20000 | 0.1158 | - | - | - | - | - |
434
- | 2.0 | 20728 | - | 0.1267 | 0.6900 (+0.1495) | 0.3621 (+0.0370) | 0.7013 (+0.2007) | 0.5844 (+0.1291) |
435
- | 2.3157 | 24000 | 0.0993 | - | - | - | - | - |
436
- | 2.7017 | 28000 | 0.0945 | - | - | - | - | - |
437
- | 3.0 | 31092 | - | 0.1184 | 0.6916 (+0.1512) | 0.3806 (+0.0556) | 0.7025 (+0.2018) | 0.5916 (+0.1362) |
438
- | -1 | -1 | - | - | 0.6902 (+0.1498) | 0.3796 (+0.0546) | 0.7323 (+0.2316) | 0.6007 (+0.1453) |
439
-
440
- * The bold row denotes the saved checkpoint.
441
 
442
  ### Framework Versions
443
  - Python: 3.11.5
 
27
  type: NanoMSMARCO_R100
28
  metrics:
29
  - type: map
30
+ value: 0.6266
31
  name: Map
32
  - type: mrr@10
33
+ value: 0.622
34
  name: Mrr@10
35
  - type: ndcg@10
36
+ value: 0.6916
37
  name: Ndcg@10
38
  - task:
39
  type: cross-encoder-reranking
 
43
  type: NanoNFCorpus_R100
44
  metrics:
45
  - type: map
46
+ value: 0.3574
47
  name: Map
48
  - type: mrr@10
49
+ value: 0.5321
50
  name: Mrr@10
51
  - type: ndcg@10
52
+ value: 0.3806
53
  name: Ndcg@10
54
  - task:
55
  type: cross-encoder-reranking
 
59
  type: NanoNQ_R100
60
  metrics:
61
  - type: map
62
+ value: 0.6437
63
  name: Map
64
  - type: mrr@10
65
+ value: 0.6663
66
  name: Mrr@10
67
  - type: ndcg@10
68
+ value: 0.7025
69
  name: Ndcg@10
70
  - task:
71
  type: cross-encoder-nano-beir
 
75
  type: NanoBEIR_R100_mean
76
  metrics:
77
  - type: map
78
+ value: 0.5426
79
  name: Map
80
  - type: mrr@10
81
+ value: 0.6068
82
  name: Mrr@10
83
  - type: ndcg@10
84
+ value: 0.5916
85
  name: Ndcg@10
86
  ---
87
 
 
191
 
192
  | Metric | NanoMSMARCO_R100 | NanoNFCorpus_R100 | NanoNQ_R100 |
193
  |:------------|:---------------------|:---------------------|:---------------------|
194
+ | map | 0.6266 (+0.1371) | 0.3574 (+0.0964) | 0.6437 (+0.2241) |
195
+ | mrr@10 | 0.6220 (+0.1445) | 0.5321 (+0.0322) | 0.6663 (+0.2396) |
196
+ | **ndcg@10** | **0.6916 (+0.1512)** | **0.3806 (+0.0556)** | **0.7025 (+0.2018)** |
197
 
198
  #### Cross Encoder Nano BEIR
199
 
 
214
 
215
  | Metric | Value |
216
  |:------------|:---------------------|
217
+ | map | 0.5426 (+0.1525) |
218
+ | mrr@10 | 0.6068 (+0.1388) |
219
+ | **ndcg@10** | **0.5916 (+0.1362)** |
220
 
221
  <!--
222
  ## Bias, Risks and Limitations
 
421
  </details>
422
 
423
  ### Training Logs
424
+ | Epoch | Step | Training Loss | Validation Loss | NanoMSMARCO_R100_ndcg@10 | NanoNFCorpus_R100_ndcg@10 | NanoNQ_R100_ndcg@10 | NanoBEIR_R100_mean_ndcg@10 |
425
+ |:------:|:-----:|:-------------:|:---------------:|:------------------------:|:-------------------------:|:-------------------:|:--------------------------:|
426
+ | -1 | -1 | - | - | 0.0186 (-0.5218) | 0.2940 (-0.0310) | 0.0428 (-0.4579) | 0.1185 (-0.3369) |
427
+ | 0.0001 | 1 | 0.7403 | - | - | - | - | - |
428
+ | 0.3860 | 4000 | 0.2255 | - | - | - | - | - |
429
+ | 0.7719 | 8000 | 0.153 | - | - | - | - | - |
430
+ | 1.0 | 10364 | - | 0.1429 | 0.6902 (+0.1498) | 0.3796 (+0.0546) | 0.7323 (+0.2316) | 0.6007 (+0.1453) |
431
+ | 1.1579 | 12000 | 0.1356 | - | - | - | - | - |
432
+ | 1.5438 | 16000 | 0.122 | - | - | - | - | - |
433
+ | 1.9298 | 20000 | 0.1158 | - | - | - | - | - |
434
+ | 2.0 | 20728 | - | 0.1267 | 0.6900 (+0.1495) | 0.3621 (+0.0370) | 0.7013 (+0.2007) | 0.5844 (+0.1291) |
435
+ | 2.3157 | 24000 | 0.0993 | - | - | - | - | - |
436
+ | 2.7017 | 28000 | 0.0945 | - | - | - | - | - |
437
+ | 3.0 | 31092 | - | 0.1184 | 0.6916 (+0.1512) | 0.3806 (+0.0556) | 0.7025 (+0.2018) | 0.5916 (+0.1362) |
438
+
 
 
439
 
440
  ### Framework Versions
441
  - Python: 3.11.5
config.json CHANGED
@@ -41,16 +41,15 @@
41
  "num_hidden_layers": 22,
42
  "pad_token_id": 50283,
43
  "position_embedding_type": "absolute",
44
- "reference_compile": true,
45
  "repad_logits_with_grad": false,
46
  "sentence_transformers": {
47
  "activation_fn": "torch.nn.modules.activation.Sigmoid",
48
- "version": "4.0.1"
49
  },
50
  "sep_token_id": 50282,
51
  "sparse_pred_ignore_index": -100,
52
  "sparse_prediction": false,
53
  "torch_dtype": "float32",
54
- "transformers_version": "4.50.3",
55
  "vocab_size": 50368
56
  }
 
41
  "num_hidden_layers": 22,
42
  "pad_token_id": 50283,
43
  "position_embedding_type": "absolute",
 
44
  "repad_logits_with_grad": false,
45
  "sentence_transformers": {
46
  "activation_fn": "torch.nn.modules.activation.Sigmoid",
47
+ "version": "4.0.2"
48
  },
49
  "sep_token_id": 50282,
50
  "sparse_pred_ignore_index": -100,
51
  "sparse_prediction": false,
52
  "torch_dtype": "float32",
53
+ "transformers_version": "4.51.0",
54
  "vocab_size": 50368
55
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f4cffca4b22f3dee1801eedc43b6390a7148e160d29c1bb412fceae36fef7f27
3
  size 598436708
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb68e7aaba743a9bafcf7fd76ee9097328acc5e550b2c18209b302ecb6ef5c8c
3
  size 598436708
tokenizer_config.json CHANGED
@@ -933,13 +933,20 @@
933
  "cls_token": "[CLS]",
934
  "extra_special_tokens": {},
935
  "mask_token": "[MASK]",
 
936
  "model_input_names": [
937
  "input_ids",
938
  "attention_mask"
939
  ],
940
  "model_max_length": 8192,
 
941
  "pad_token": "[PAD]",
 
 
942
  "sep_token": "[SEP]",
 
943
  "tokenizer_class": "PreTrainedTokenizer",
 
 
944
  "unk_token": "[UNK]"
945
  }
 
933
  "cls_token": "[CLS]",
934
  "extra_special_tokens": {},
935
  "mask_token": "[MASK]",
936
+ "max_length": 8192,
937
  "model_input_names": [
938
  "input_ids",
939
  "attention_mask"
940
  ],
941
  "model_max_length": 8192,
942
+ "pad_to_multiple_of": null,
943
  "pad_token": "[PAD]",
944
+ "pad_token_type_id": 0,
945
+ "padding_side": "right",
946
  "sep_token": "[SEP]",
947
+ "stride": 0,
948
  "tokenizer_class": "PreTrainedTokenizer",
949
+ "truncation_side": "right",
950
+ "truncation_strategy": "longest_first",
951
  "unk_token": "[UNK]"
952
  }