Minbyul commited on
Commit
dfe4fcc
1 Parent(s): 8af1b21

Model save

Browse files
Files changed (6) hide show
  1. README.md +12 -17
  2. all_results.json +4 -4
  3. config.json +1 -1
  4. train_results.json +4 -4
  5. trainer_state.json +4 -4
  6. training_args.bin +1 -1
README.md CHANGED
@@ -2,15 +2,10 @@
2
  license: llama2
3
  base_model: epfl-llm/meditron-7b
4
  tags:
5
- - alignment-handbook
6
- - trl
7
- - dpo
8
- - generated_from_trainer
9
  - trl
10
  - dpo
 
11
  - generated_from_trainer
12
- datasets:
13
- - HuggingFaceH4/ultrafeedback_binarized
14
  model-index:
15
  - name: meditron-7b-dpo-full-wo-kqa_golden-ep3
16
  results: []
@@ -21,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
21
 
22
  # meditron-7b-dpo-full-wo-kqa_golden-ep3
23
 
24
- This model is a fine-tuned version of [epfl-llm/meditron-7b](https://huggingface.co/epfl-llm/meditron-7b) on the HuggingFaceH4/ultrafeedback_binarized dataset.
25
  It achieves the following results on the evaluation set:
 
 
 
 
26
  - Loss: 0.4459
27
- - Rewards/chosen: -0.4566
28
- - Rewards/rejected: -1.4012
29
  - Rewards/accuracies: 0.8068
 
30
  - Rewards/margins: 0.9447
31
- - Logps/rejected: -1444.6896
32
- - Logps/chosen: -859.0582
33
- - Logits/rejected: -0.9203
34
- - Logits/chosen: -0.8310
35
 
36
  ## Model description
37
 
@@ -66,10 +61,10 @@ The following hyperparameters were used during training:
66
 
67
  ### Training results
68
 
69
- | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
70
- |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
71
- | 0.5643 | 0.5 | 100 | 0.5890 | -0.0484 | -0.2951 | 0.7727 | 0.2467 | -1334.0771 | -818.2397 | -0.8645 | -0.6995 |
72
- | 0.3959 | 1.0 | 200 | 0.4459 | -0.4566 | -1.4012 | 0.8068 | 0.9447 | -1444.6896 | -859.0582 | -0.9203 | -0.8310 |
73
 
74
 
75
  ### Framework versions
 
2
  license: llama2
3
  base_model: epfl-llm/meditron-7b
4
  tags:
 
 
 
 
5
  - trl
6
  - dpo
7
+ - alignment-handbook
8
  - generated_from_trainer
 
 
9
  model-index:
10
  - name: meditron-7b-dpo-full-wo-kqa_golden-ep3
11
  results: []
 
16
 
17
  # meditron-7b-dpo-full-wo-kqa_golden-ep3
18
 
19
+ This model is a fine-tuned version of [epfl-llm/meditron-7b](https://huggingface.co/epfl-llm/meditron-7b) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
+ - Logits/chosen: -0.8310
22
+ - Logits/rejected: -0.9203
23
+ - Logps/chosen: -859.0582
24
+ - Logps/rejected: -1444.6896
25
  - Loss: 0.4459
 
 
26
  - Rewards/accuracies: 0.8068
27
+ - Rewards/chosen: -0.4566
28
  - Rewards/margins: 0.9447
29
+ - Rewards/rejected: -1.4012
 
 
 
30
 
31
  ## Model description
32
 
 
61
 
62
  ### Training results
63
 
64
+ | Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
65
+ |:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
66
+ | 0.5643 | 0.5 | 100 | -0.6995 | -0.8645 | -818.2397 | -1334.0771 | 0.5890 | 0.7727 | -0.0484 | 0.2467 | -0.2951 |
67
+ | 0.3959 | 1.0 | 200 | -0.8310 | -0.9203 | -859.0582 | -1444.6896 | 0.4459 | 0.8068 | -0.4566 | 0.9447 | -1.4012 |
68
 
69
 
70
  ### Framework versions
all_results.json CHANGED
@@ -13,9 +13,9 @@
13
  "eval_samples": 676,
14
  "eval_samples_per_second": 8.395,
15
  "eval_steps_per_second": 0.273,
16
- "train_loss": 0.5449820566177368,
17
- "train_runtime": 3295.9797,
18
  "train_samples": 12791,
19
- "train_samples_per_second": 3.881,
20
- "train_steps_per_second": 0.061
21
  }
 
13
  "eval_samples": 676,
14
  "eval_samples_per_second": 8.395,
15
  "eval_steps_per_second": 0.273,
16
+ "train_loss": 0.0,
17
+ "train_runtime": 2.1286,
18
  "train_samples": 12791,
19
+ "train_samples_per_second": 6009.015,
20
+ "train_steps_per_second": 93.957
21
  }
config.json CHANGED
@@ -23,6 +23,6 @@
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.39.0.dev0",
26
- "use_cache": true,
27
  "vocab_size": 32017
28
  }
 
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.39.0.dev0",
26
+ "use_cache": false,
27
  "vocab_size": 32017
28
  }
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 1.0,
3
- "train_loss": 0.5449820566177368,
4
- "train_runtime": 3295.9797,
5
  "train_samples": 12791,
6
- "train_samples_per_second": 3.881,
7
- "train_steps_per_second": 0.061
8
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "train_loss": 0.0,
4
+ "train_runtime": 2.1286,
5
  "train_samples": 12791,
6
+ "train_samples_per_second": 6009.015,
7
+ "train_steps_per_second": 93.957
8
  }
trainer_state.json CHANGED
@@ -359,10 +359,10 @@
359
  "epoch": 1.0,
360
  "step": 200,
361
  "total_flos": 0.0,
362
- "train_loss": 0.5449820566177368,
363
- "train_runtime": 3295.9797,
364
- "train_samples_per_second": 3.881,
365
- "train_steps_per_second": 0.061
366
  }
367
  ],
368
  "logging_steps": 10,
 
359
  "epoch": 1.0,
360
  "step": 200,
361
  "total_flos": 0.0,
362
+ "train_loss": 0.0,
363
+ "train_runtime": 2.1286,
364
+ "train_samples_per_second": 6009.015,
365
+ "train_steps_per_second": 93.957
366
  }
367
  ],
368
  "logging_steps": 10,
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dd53dda42c7e5233cf878ab56800a1d8b2bf9a44cdfc7c732ed3fd104c7948ce
3
  size 6264
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:beea2194ed996027ed8a95f01bc25fff671b87a7c15e301980832509c5fe90a9
3
  size 6264