dejori commited on
Commit
ddd8741
·
verified ·
1 Parent(s): 9de498f

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +8 -0
  2. gemma-9b-dpo/README.md +72 -0
  3. gemma-9b-dpo/adapter_config.json +5 -5
  4. gemma-9b-dpo/adapter_model.safetensors +1 -1
  5. gemma-9b-dpo/checkpoint-100/README.md +209 -0
  6. gemma-9b-dpo/checkpoint-100/adapter_config.json +46 -0
  7. gemma-9b-dpo/checkpoint-100/adapter_model.safetensors +3 -0
  8. gemma-9b-dpo/checkpoint-100/chat_template.jinja +4 -0
  9. gemma-9b-dpo/checkpoint-100/optimizer.pt +3 -0
  10. gemma-9b-dpo/checkpoint-100/rng_state.pth +3 -0
  11. gemma-9b-dpo/checkpoint-100/scheduler.pt +3 -0
  12. gemma-9b-dpo/checkpoint-100/tokenizer.json +3 -0
  13. gemma-9b-dpo/checkpoint-100/tokenizer_config.json +19 -0
  14. gemma-9b-dpo/checkpoint-100/trainer_state.json +334 -0
  15. gemma-9b-dpo/checkpoint-100/training_args.bin +3 -0
  16. gemma-9b-dpo/checkpoint-150/README.md +209 -0
  17. gemma-9b-dpo/checkpoint-150/adapter_config.json +46 -0
  18. gemma-9b-dpo/checkpoint-150/adapter_model.safetensors +3 -0
  19. gemma-9b-dpo/checkpoint-150/chat_template.jinja +4 -0
  20. gemma-9b-dpo/checkpoint-150/optimizer.pt +3 -0
  21. gemma-9b-dpo/checkpoint-150/rng_state.pth +3 -0
  22. gemma-9b-dpo/checkpoint-150/scheduler.pt +3 -0
  23. gemma-9b-dpo/checkpoint-150/tokenizer.json +3 -0
  24. gemma-9b-dpo/checkpoint-150/tokenizer_config.json +19 -0
  25. gemma-9b-dpo/checkpoint-150/trainer_state.json +484 -0
  26. gemma-9b-dpo/checkpoint-150/training_args.bin +3 -0
  27. gemma-9b-dpo/checkpoint-200/README.md +209 -0
  28. gemma-9b-dpo/checkpoint-200/adapter_config.json +46 -0
  29. gemma-9b-dpo/checkpoint-200/adapter_model.safetensors +3 -0
  30. gemma-9b-dpo/checkpoint-200/chat_template.jinja +4 -0
  31. gemma-9b-dpo/checkpoint-200/optimizer.pt +3 -0
  32. gemma-9b-dpo/checkpoint-200/rng_state.pth +3 -0
  33. gemma-9b-dpo/checkpoint-200/scheduler.pt +3 -0
  34. gemma-9b-dpo/checkpoint-200/tokenizer.json +3 -0
  35. gemma-9b-dpo/checkpoint-200/tokenizer_config.json +19 -0
  36. gemma-9b-dpo/checkpoint-200/trainer_state.json +634 -0
  37. gemma-9b-dpo/checkpoint-200/training_args.bin +3 -0
  38. gemma-9b-dpo/checkpoint-250/README.md +209 -0
  39. gemma-9b-dpo/checkpoint-250/adapter_config.json +46 -0
  40. gemma-9b-dpo/checkpoint-250/adapter_model.safetensors +3 -0
  41. gemma-9b-dpo/checkpoint-250/chat_template.jinja +4 -0
  42. gemma-9b-dpo/checkpoint-250/optimizer.pt +3 -0
  43. gemma-9b-dpo/checkpoint-250/rng_state.pth +3 -0
  44. gemma-9b-dpo/checkpoint-250/scheduler.pt +3 -0
  45. gemma-9b-dpo/checkpoint-250/tokenizer.json +3 -0
  46. gemma-9b-dpo/checkpoint-250/tokenizer_config.json +19 -0
  47. gemma-9b-dpo/checkpoint-250/trainer_state.json +784 -0
  48. gemma-9b-dpo/checkpoint-250/training_args.bin +3 -0
  49. gemma-9b-dpo/checkpoint-300/README.md +209 -0
  50. gemma-9b-dpo/checkpoint-300/adapter_config.json +46 -0
.gitattributes CHANGED
@@ -52,3 +52,11 @@ gemma-2b-dpo/checkpoint-540/tokenizer.json filter=lfs diff=lfs merge=lfs -text
52
  gemma-2b-distilled/checkpoint-150/tokenizer.json filter=lfs diff=lfs merge=lfs -text
53
  gemma-2b-distilled/checkpoint-225/tokenizer.json filter=lfs diff=lfs merge=lfs -text
54
  gemma-2b-distilled/checkpoint-75/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
52
  gemma-2b-distilled/checkpoint-150/tokenizer.json filter=lfs diff=lfs merge=lfs -text
53
  gemma-2b-distilled/checkpoint-225/tokenizer.json filter=lfs diff=lfs merge=lfs -text
54
  gemma-2b-distilled/checkpoint-75/tokenizer.json filter=lfs diff=lfs merge=lfs -text
55
+ gemma-9b-dpo/checkpoint-100/tokenizer.json filter=lfs diff=lfs merge=lfs -text
56
+ gemma-9b-dpo/checkpoint-150/tokenizer.json filter=lfs diff=lfs merge=lfs -text
57
+ gemma-9b-dpo/checkpoint-200/tokenizer.json filter=lfs diff=lfs merge=lfs -text
58
+ gemma-9b-dpo/checkpoint-250/tokenizer.json filter=lfs diff=lfs merge=lfs -text
59
+ gemma-9b-dpo/checkpoint-300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
60
+ gemma-9b-dpo/checkpoint-350/tokenizer.json filter=lfs diff=lfs merge=lfs -text
61
+ gemma-9b-dpo/checkpoint-351/tokenizer.json filter=lfs diff=lfs merge=lfs -text
62
+ gemma-9b-dpo/checkpoint-50/tokenizer.json filter=lfs diff=lfs merge=lfs -text
gemma-9b-dpo/README.md ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: google/gemma-2-9b-it
3
+ library_name: peft
4
+ model_name: gemma-9b-dpo-medgemma-450
5
+ tags:
6
+ - base_model:adapter:google/gemma-2-9b-it
7
+ - dpo
8
+ - lora
9
+ - transformers
10
+ - trl
11
+ licence: license
12
+ pipeline_tag: text-generation
13
+ ---
14
+
15
+ # Model Card for gemma-9b-dpo-medgemma-450
16
+
17
+ This model is a fine-tuned version of [google/gemma-2-9b-it](https://huggingface.co/google/gemma-2-9b-it).
18
+ It has been trained using [TRL](https://github.com/huggingface/trl).
19
+
20
+ ## Quick start
21
+
22
+ ```python
23
+ from transformers import pipeline
24
+
25
+ question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
26
+ generator = pipeline("text-generation", model="None", device="cuda")
27
+ output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
28
+ print(output["generated_text"])
29
+ ```
30
+
31
+ ## Training procedure
32
+
33
+
34
+
35
+
36
+ This model was trained with DPO, a method introduced in [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290).
37
+
38
+ ### Framework versions
39
+
40
+ - PEFT 0.18.1
41
+ - TRL: 0.28.0
42
+ - Transformers: 5.2.0
43
+ - Pytorch: 2.10.0
44
+ - Datasets: 4.5.0
45
+ - Tokenizers: 0.22.2
46
+
47
+ ## Citations
48
+
49
+ Cite DPO as:
50
+
51
+ ```bibtex
52
+ @inproceedings{rafailov2023direct,
53
+ title = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
54
+ author = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
55
+ year = 2023,
56
+ booktitle = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
57
+ url = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
58
+ editor = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
59
+ }
60
+ ```
61
+
62
+ Cite TRL as:
63
+
64
+ ```bibtex
65
+ @software{vonwerra2020trl,
66
+ title = {{TRL: Transformers Reinforcement Learning}},
67
+ author = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
68
+ license = {Apache-2.0},
69
+ url = {https://github.com/huggingface/trl},
70
+ year = {2020}
71
+ }
72
+ ```
gemma-9b-dpo/adapter_config.json CHANGED
@@ -29,13 +29,13 @@
29
  "rank_pattern": {},
30
  "revision": null,
31
  "target_modules": [
32
- "gate_proj",
33
- "q_proj",
34
  "v_proj",
35
- "up_proj",
36
- "o_proj",
37
  "k_proj",
38
- "down_proj"
 
 
 
39
  ],
40
  "target_parameters": null,
41
  "task_type": "CAUSAL_LM",
 
29
  "rank_pattern": {},
30
  "revision": null,
31
  "target_modules": [
32
+ "down_proj",
 
33
  "v_proj",
 
 
34
  "k_proj",
35
+ "up_proj",
36
+ "q_proj",
37
+ "gate_proj",
38
+ "o_proj"
39
  ],
40
  "target_parameters": null,
41
  "task_type": "CAUSAL_LM",
gemma-9b-dpo/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b999ce33083a55536a152a5a3c149746e107f5d28f96abd7f5750c9bf520bed6
3
  size 216151256
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ea0ff88e00dfb29580d0fa61936b45ac7ee6e5f886f57fd984f2f3854211d5e
3
  size 216151256
gemma-9b-dpo/checkpoint-100/README.md ADDED
@@ -0,0 +1,209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: google/gemma-2-9b-it
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - base_model:adapter:google/gemma-2-9b-it
7
+ - dpo
8
+ - lora
9
+ - transformers
10
+ - trl
11
+ ---
12
+
13
+ # Model Card for Model ID
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+
19
+ ## Model Details
20
+
21
+ ### Model Description
22
+
23
+ <!-- Provide a longer summary of what this model is. -->
24
+
25
+
26
+
27
+ - **Developed by:** [More Information Needed]
28
+ - **Funded by [optional]:** [More Information Needed]
29
+ - **Shared by [optional]:** [More Information Needed]
30
+ - **Model type:** [More Information Needed]
31
+ - **Language(s) (NLP):** [More Information Needed]
32
+ - **License:** [More Information Needed]
33
+ - **Finetuned from model [optional]:** [More Information Needed]
34
+
35
+ ### Model Sources [optional]
36
+
37
+ <!-- Provide the basic links for the model. -->
38
+
39
+ - **Repository:** [More Information Needed]
40
+ - **Paper [optional]:** [More Information Needed]
41
+ - **Demo [optional]:** [More Information Needed]
42
+
43
+ ## Uses
44
+
45
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
46
+
47
+ ### Direct Use
48
+
49
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
50
+
51
+ [More Information Needed]
52
+
53
+ ### Downstream Use [optional]
54
+
55
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
56
+
57
+ [More Information Needed]
58
+
59
+ ### Out-of-Scope Use
60
+
61
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
62
+
63
+ [More Information Needed]
64
+
65
+ ## Bias, Risks, and Limitations
66
+
67
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
68
+
69
+ [More Information Needed]
70
+
71
+ ### Recommendations
72
+
73
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
74
+
75
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
76
+
77
+ ## How to Get Started with the Model
78
+
79
+ Use the code below to get started with the model.
80
+
81
+ [More Information Needed]
82
+
83
+ ## Training Details
84
+
85
+ ### Training Data
86
+
87
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
88
+
89
+ [More Information Needed]
90
+
91
+ ### Training Procedure
92
+
93
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
94
+
95
+ #### Preprocessing [optional]
96
+
97
+ [More Information Needed]
98
+
99
+
100
+ #### Training Hyperparameters
101
+
102
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
103
+
104
+ #### Speeds, Sizes, Times [optional]
105
+
106
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
107
+
108
+ [More Information Needed]
109
+
110
+ ## Evaluation
111
+
112
+ <!-- This section describes the evaluation protocols and provides the results. -->
113
+
114
+ ### Testing Data, Factors & Metrics
115
+
116
+ #### Testing Data
117
+
118
+ <!-- This should link to a Dataset Card if possible. -->
119
+
120
+ [More Information Needed]
121
+
122
+ #### Factors
123
+
124
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
125
+
126
+ [More Information Needed]
127
+
128
+ #### Metrics
129
+
130
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
131
+
132
+ [More Information Needed]
133
+
134
+ ### Results
135
+
136
+ [More Information Needed]
137
+
138
+ #### Summary
139
+
140
+
141
+
142
+ ## Model Examination [optional]
143
+
144
+ <!-- Relevant interpretability work for the model goes here -->
145
+
146
+ [More Information Needed]
147
+
148
+ ## Environmental Impact
149
+
150
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
151
+
152
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
153
+
154
+ - **Hardware Type:** [More Information Needed]
155
+ - **Hours used:** [More Information Needed]
156
+ - **Cloud Provider:** [More Information Needed]
157
+ - **Compute Region:** [More Information Needed]
158
+ - **Carbon Emitted:** [More Information Needed]
159
+
160
+ ## Technical Specifications [optional]
161
+
162
+ ### Model Architecture and Objective
163
+
164
+ [More Information Needed]
165
+
166
+ ### Compute Infrastructure
167
+
168
+ [More Information Needed]
169
+
170
+ #### Hardware
171
+
172
+ [More Information Needed]
173
+
174
+ #### Software
175
+
176
+ [More Information Needed]
177
+
178
+ ## Citation [optional]
179
+
180
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
181
+
182
+ **BibTeX:**
183
+
184
+ [More Information Needed]
185
+
186
+ **APA:**
187
+
188
+ [More Information Needed]
189
+
190
+ ## Glossary [optional]
191
+
192
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
193
+
194
+ [More Information Needed]
195
+
196
+ ## More Information [optional]
197
+
198
+ [More Information Needed]
199
+
200
+ ## Model Card Authors [optional]
201
+
202
+ [More Information Needed]
203
+
204
+ ## Model Card Contact
205
+
206
+ [More Information Needed]
207
+ ### Framework versions
208
+
209
+ - PEFT 0.18.1
gemma-9b-dpo/checkpoint-100/adapter_config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "google/gemma-2-9b-it",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": false,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 32,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.05,
22
+ "megatron_config": null,
23
+ "megatron_core": "megatron.core",
24
+ "modules_to_save": null,
25
+ "peft_type": "LORA",
26
+ "peft_version": "0.18.1",
27
+ "qalora_group_size": 16,
28
+ "r": 16,
29
+ "rank_pattern": {},
30
+ "revision": null,
31
+ "target_modules": [
32
+ "down_proj",
33
+ "v_proj",
34
+ "k_proj",
35
+ "up_proj",
36
+ "q_proj",
37
+ "gate_proj",
38
+ "o_proj"
39
+ ],
40
+ "target_parameters": null,
41
+ "task_type": "CAUSAL_LM",
42
+ "trainable_token_indices": null,
43
+ "use_dora": false,
44
+ "use_qalora": false,
45
+ "use_rslora": false
46
+ }
gemma-9b-dpo/checkpoint-100/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7c9f0c8577a228cc7034adc448f4596f9de36d2fa777c37a8d89970d6026ef2
3
+ size 216151256
gemma-9b-dpo/checkpoint-100/chat_template.jinja ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {{ bos_token }}{% if messages[0]['role'] == 'system' %}{{ raise_exception('System role not supported') }}{% endif %}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if (message['role'] == 'assistant') %}{% set role = 'model' %}{% else %}{% set role = message['role'] %}{% endif %}{{ '<start_of_turn>' + role + '
2
+ ' + message['content'] | trim + '<end_of_turn>
3
+ ' }}{% endfor %}{% if add_generation_prompt %}{{'<start_of_turn>model
4
+ '}}{% endif %}
gemma-9b-dpo/checkpoint-100/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf0ce732e832a1274480b5520b72d3661260561069eaf76121fe58f232ed9d84
3
+ size 110425877
gemma-9b-dpo/checkpoint-100/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:250560ab3d528161ab3659b120def6e4a9ab4b457e3399603bbcfa40db3efc90
3
+ size 14645
gemma-9b-dpo/checkpoint-100/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67328bf66accf20808215e57905f6c9930b4b18ece0e4f4e03baee7eb2fb81fa
3
+ size 1465
gemma-9b-dpo/checkpoint-100/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:394ace002a144ac6ad5486387502f2d36f70c087310c3d907857240c76fcb36e
3
+ size 34362748
gemma-9b-dpo/checkpoint-100/tokenizer_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "<bos>",
4
+ "clean_up_tokenization_spaces": false,
5
+ "eos_token": "<eos>",
6
+ "extra_special_tokens": [
7
+ "<start_of_turn>",
8
+ "<end_of_turn>"
9
+ ],
10
+ "is_local": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 1000000000000000019884624838656,
13
+ "pad_token": "<pad>",
14
+ "sp_model_kwargs": {},
15
+ "spaces_between_special_tokens": false,
16
+ "tokenizer_class": "GemmaTokenizer",
17
+ "unk_token": "<unk>",
18
+ "use_default_system_prompt": false
19
+ }
gemma-9b-dpo/checkpoint-100/trainer_state.json ADDED
@@ -0,0 +1,334 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.8602150537634409,
6
+ "eval_steps": 500,
7
+ "global_step": 100,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.043010752688172046,
14
+ "grad_norm": 1.6562875509262085,
15
+ "learning_rate": 3.333333333333333e-07,
16
+ "logits/chosen": -5.94815731048584,
17
+ "logits/rejected": -5.856410503387451,
18
+ "logps/chosen": -302.07794189453125,
19
+ "logps/rejected": -283.5802001953125,
20
+ "loss": 0.6852486610412598,
21
+ "rewards/accuracies": 0.375,
22
+ "rewards/chosen": 0.01349079143255949,
23
+ "rewards/margins": 0.017757166177034378,
24
+ "rewards/rejected": -0.00426637614145875,
25
+ "step": 5
26
+ },
27
+ {
28
+ "epoch": 0.08602150537634409,
29
+ "grad_norm": 1.7723957300186157,
30
+ "learning_rate": 7.5e-07,
31
+ "logits/chosen": -5.585428714752197,
32
+ "logits/rejected": -5.577895641326904,
33
+ "logps/chosen": -362.89239501953125,
34
+ "logps/rejected": -280.8010559082031,
35
+ "loss": 0.6844330310821534,
36
+ "rewards/accuracies": 0.6499999761581421,
37
+ "rewards/chosen": 0.018745478242635727,
38
+ "rewards/margins": 0.018739622086286545,
39
+ "rewards/rejected": 5.85438692723983e-06,
40
+ "step": 10
41
+ },
42
+ {
43
+ "epoch": 0.12903225806451613,
44
+ "grad_norm": 1.6630758047103882,
45
+ "learning_rate": 1.1666666666666668e-06,
46
+ "logits/chosen": -5.916059494018555,
47
+ "logits/rejected": -5.797668933868408,
48
+ "logps/chosen": -307.4602966308594,
49
+ "logps/rejected": -316.79803466796875,
50
+ "loss": 0.7070020198822021,
51
+ "rewards/accuracies": 0.5249999761581421,
52
+ "rewards/chosen": -0.008706416003406048,
53
+ "rewards/margins": -0.02138950116932392,
54
+ "rewards/rejected": 0.012683087959885597,
55
+ "step": 15
56
+ },
57
+ {
58
+ "epoch": 0.17204301075268819,
59
+ "grad_norm": 5.225494861602783,
60
+ "learning_rate": 1.5833333333333333e-06,
61
+ "logits/chosen": -5.818185329437256,
62
+ "logits/rejected": -5.638014793395996,
63
+ "logps/chosen": -329.3303527832031,
64
+ "logps/rejected": -292.2645568847656,
65
+ "loss": 0.6929943561553955,
66
+ "rewards/accuracies": 0.4749999940395355,
67
+ "rewards/chosen": -0.044999174773693085,
68
+ "rewards/margins": 0.00216915225610137,
69
+ "rewards/rejected": -0.04716832935810089,
70
+ "step": 20
71
+ },
72
+ {
73
+ "epoch": 0.21505376344086022,
74
+ "grad_norm": 2.5229244232177734,
75
+ "learning_rate": 2e-06,
76
+ "logits/chosen": -5.8189592361450195,
77
+ "logits/rejected": -5.646462917327881,
78
+ "logps/chosen": -311.9922180175781,
79
+ "logps/rejected": -297.0098571777344,
80
+ "loss": 0.7002316951751709,
81
+ "rewards/accuracies": 0.4000000059604645,
82
+ "rewards/chosen": 0.0544702522456646,
83
+ "rewards/margins": -0.012132339179515839,
84
+ "rewards/rejected": 0.06660258769989014,
85
+ "step": 25
86
+ },
87
+ {
88
+ "epoch": 0.25806451612903225,
89
+ "grad_norm": 1.6524324417114258,
90
+ "learning_rate": 2.4166666666666667e-06,
91
+ "logits/chosen": -5.696784496307373,
92
+ "logits/rejected": -5.6943678855896,
93
+ "logps/chosen": -350.2938537597656,
94
+ "logps/rejected": -322.43182373046875,
95
+ "loss": 0.699475908279419,
96
+ "rewards/accuracies": 0.574999988079071,
97
+ "rewards/chosen": -0.0020792910363525152,
98
+ "rewards/margins": 0.02348965033888817,
99
+ "rewards/rejected": -0.025568943470716476,
100
+ "step": 30
101
+ },
102
+ {
103
+ "epoch": 0.3010752688172043,
104
+ "grad_norm": 9.075493812561035,
105
+ "learning_rate": 2.8333333333333335e-06,
106
+ "logits/chosen": -5.407717704772949,
107
+ "logits/rejected": -5.5642805099487305,
108
+ "logps/chosen": -391.3837585449219,
109
+ "logps/rejected": -333.9232482910156,
110
+ "loss": 0.6926415920257568,
111
+ "rewards/accuracies": 0.4749999940395355,
112
+ "rewards/chosen": -0.026233959943056107,
113
+ "rewards/margins": 0.00582819152623415,
114
+ "rewards/rejected": -0.03206215053796768,
115
+ "step": 35
116
+ },
117
+ {
118
+ "epoch": 0.34408602150537637,
119
+ "grad_norm": 2.6282100677490234,
120
+ "learning_rate": 2.9714285714285716e-06,
121
+ "logits/chosen": -5.649778366088867,
122
+ "logits/rejected": -5.512002944946289,
123
+ "logps/chosen": -338.31048583984375,
124
+ "logps/rejected": -279.06536865234375,
125
+ "loss": 0.6739111423492432,
126
+ "rewards/accuracies": 0.625,
127
+ "rewards/chosen": 0.04578382521867752,
128
+ "rewards/margins": 0.042618148028850555,
129
+ "rewards/rejected": 0.0031656839419156313,
130
+ "step": 40
131
+ },
132
+ {
133
+ "epoch": 0.3870967741935484,
134
+ "grad_norm": 16.96689796447754,
135
+ "learning_rate": 2.923809523809524e-06,
136
+ "logits/chosen": -5.518254280090332,
137
+ "logits/rejected": -5.577255725860596,
138
+ "logps/chosen": -374.29449462890625,
139
+ "logps/rejected": -366.4861145019531,
140
+ "loss": 0.7162120819091797,
141
+ "rewards/accuracies": 0.550000011920929,
142
+ "rewards/chosen": 0.03416462987661362,
143
+ "rewards/margins": 0.0036566159687936306,
144
+ "rewards/rejected": 0.030508000403642654,
145
+ "step": 45
146
+ },
147
+ {
148
+ "epoch": 0.43010752688172044,
149
+ "grad_norm": 1.447816252708435,
150
+ "learning_rate": 2.8761904761904764e-06,
151
+ "logits/chosen": -5.831571102142334,
152
+ "logits/rejected": -5.802765846252441,
153
+ "logps/chosen": -316.0965881347656,
154
+ "logps/rejected": -269.0494384765625,
155
+ "loss": 0.6838803291320801,
156
+ "rewards/accuracies": 0.550000011920929,
157
+ "rewards/chosen": -0.1265859305858612,
158
+ "rewards/margins": 0.026753634214401245,
159
+ "rewards/rejected": -0.15333956480026245,
160
+ "step": 50
161
+ },
162
+ {
163
+ "epoch": 0.4731182795698925,
164
+ "grad_norm": 2.174459457397461,
165
+ "learning_rate": 2.8285714285714288e-06,
166
+ "logits/chosen": -5.777364253997803,
167
+ "logits/rejected": -5.834509372711182,
168
+ "logps/chosen": -316.31390380859375,
169
+ "logps/rejected": -282.5376892089844,
170
+ "loss": 0.6832016944885254,
171
+ "rewards/accuracies": 0.675000011920929,
172
+ "rewards/chosen": -0.01756533980369568,
173
+ "rewards/margins": 0.02713041380047798,
174
+ "rewards/rejected": -0.04469575732946396,
175
+ "step": 55
176
+ },
177
+ {
178
+ "epoch": 0.5161290322580645,
179
+ "grad_norm": 2.2876009941101074,
180
+ "learning_rate": 2.780952380952381e-06,
181
+ "logits/chosen": -5.92340087890625,
182
+ "logits/rejected": -5.846226692199707,
183
+ "logps/chosen": -235.49667358398438,
184
+ "logps/rejected": -267.7764892578125,
185
+ "loss": 0.6656109809875488,
186
+ "rewards/accuracies": 0.6499999761581421,
187
+ "rewards/chosen": 0.021378064528107643,
188
+ "rewards/margins": 0.06001085042953491,
189
+ "rewards/rejected": -0.038632798939943314,
190
+ "step": 60
191
+ },
192
+ {
193
+ "epoch": 0.5591397849462365,
194
+ "grad_norm": 3.45564603805542,
195
+ "learning_rate": 2.733333333333333e-06,
196
+ "logits/chosen": -5.650258541107178,
197
+ "logits/rejected": -5.6721577644348145,
198
+ "logps/chosen": -381.40655517578125,
199
+ "logps/rejected": -311.68865966796875,
200
+ "loss": 0.7090614795684814,
201
+ "rewards/accuracies": 0.4749999940395355,
202
+ "rewards/chosen": 0.1167660504579544,
203
+ "rewards/margins": -0.0002812861348502338,
204
+ "rewards/rejected": 0.11704733222723007,
205
+ "step": 65
206
+ },
207
+ {
208
+ "epoch": 0.6021505376344086,
209
+ "grad_norm": 4.502689838409424,
210
+ "learning_rate": 2.685714285714286e-06,
211
+ "logits/chosen": -5.727735996246338,
212
+ "logits/rejected": -5.644078254699707,
213
+ "logps/chosen": -362.89886474609375,
214
+ "logps/rejected": -314.7126770019531,
215
+ "loss": 0.6596640110015869,
216
+ "rewards/accuracies": 0.6499999761581421,
217
+ "rewards/chosen": 0.07285688072443008,
218
+ "rewards/margins": 0.07270809262990952,
219
+ "rewards/rejected": 0.00014879256195854396,
220
+ "step": 70
221
+ },
222
+ {
223
+ "epoch": 0.6451612903225806,
224
+ "grad_norm": 2.2681005001068115,
225
+ "learning_rate": 2.638095238095238e-06,
226
+ "logits/chosen": -5.528594017028809,
227
+ "logits/rejected": -5.29849910736084,
228
+ "logps/chosen": -353.0726013183594,
229
+ "logps/rejected": -331.8568115234375,
230
+ "loss": 0.7387234687805175,
231
+ "rewards/accuracies": 0.574999988079071,
232
+ "rewards/chosen": 0.1685911864042282,
233
+ "rewards/margins": -0.04706493765115738,
234
+ "rewards/rejected": 0.2156561315059662,
235
+ "step": 75
236
+ },
237
+ {
238
+ "epoch": 0.6881720430107527,
239
+ "grad_norm": 3.0135886669158936,
240
+ "learning_rate": 2.5904761904761907e-06,
241
+ "logits/chosen": -5.8677144050598145,
242
+ "logits/rejected": -5.5384016036987305,
243
+ "logps/chosen": -326.17724609375,
244
+ "logps/rejected": -285.8974304199219,
245
+ "loss": 0.6047093391418457,
246
+ "rewards/accuracies": 0.925000011920929,
247
+ "rewards/chosen": 0.19864055514335632,
248
+ "rewards/margins": 0.19678126275539398,
249
+ "rewards/rejected": 0.0018593042623251677,
250
+ "step": 80
251
+ },
252
+ {
253
+ "epoch": 0.7311827956989247,
254
+ "grad_norm": 1.7684988975524902,
255
+ "learning_rate": 2.5428571428571427e-06,
256
+ "logits/chosen": -5.413943290710449,
257
+ "logits/rejected": -5.427316665649414,
258
+ "logps/chosen": -372.1170654296875,
259
+ "logps/rejected": -312.30279541015625,
260
+ "loss": 0.651799201965332,
261
+ "rewards/accuracies": 0.675000011920929,
262
+ "rewards/chosen": 0.179380863904953,
263
+ "rewards/margins": 0.09846550226211548,
264
+ "rewards/rejected": 0.08091535419225693,
265
+ "step": 85
266
+ },
267
+ {
268
+ "epoch": 0.7741935483870968,
269
+ "grad_norm": 2.4072341918945312,
270
+ "learning_rate": 2.4952380952380955e-06,
271
+ "logits/chosen": -5.4685492515563965,
272
+ "logits/rejected": -5.51275634765625,
273
+ "logps/chosen": -345.37188720703125,
274
+ "logps/rejected": -273.87261962890625,
275
+ "loss": 0.6362700462341309,
276
+ "rewards/accuracies": 0.6499999761581421,
277
+ "rewards/chosen": 0.33122482895851135,
278
+ "rewards/margins": 0.1338074505329132,
279
+ "rewards/rejected": 0.19741736352443695,
280
+ "step": 90
281
+ },
282
+ {
283
+ "epoch": 0.8172043010752689,
284
+ "grad_norm": 1.745698094367981,
285
+ "learning_rate": 2.4476190476190475e-06,
286
+ "logits/chosen": -5.6481757164001465,
287
+ "logits/rejected": -5.4260029792785645,
288
+ "logps/chosen": -337.11767578125,
289
+ "logps/rejected": -263.3458251953125,
290
+ "loss": 0.6222721099853515,
291
+ "rewards/accuracies": 0.824999988079071,
292
+ "rewards/chosen": 0.25297099351882935,
293
+ "rewards/margins": 0.1701255589723587,
294
+ "rewards/rejected": 0.08284540474414825,
295
+ "step": 95
296
+ },
297
+ {
298
+ "epoch": 0.8602150537634409,
299
+ "grad_norm": 1.4592756032943726,
300
+ "learning_rate": 2.4000000000000003e-06,
301
+ "logits/chosen": -5.402789115905762,
302
+ "logits/rejected": -5.280846118927002,
303
+ "logps/chosen": -304.55316162109375,
304
+ "logps/rejected": -314.95306396484375,
305
+ "loss": 0.6404934883117676,
306
+ "rewards/accuracies": 0.699999988079071,
307
+ "rewards/chosen": 0.29103565216064453,
308
+ "rewards/margins": 0.12328717857599258,
309
+ "rewards/rejected": 0.16774848103523254,
310
+ "step": 100
311
+ }
312
+ ],
313
+ "logging_steps": 5,
314
+ "max_steps": 351,
315
+ "num_input_tokens_seen": 0,
316
+ "num_train_epochs": 3,
317
+ "save_steps": 50,
318
+ "stateful_callbacks": {
319
+ "TrainerControl": {
320
+ "args": {
321
+ "should_epoch_stop": false,
322
+ "should_evaluate": false,
323
+ "should_log": false,
324
+ "should_save": true,
325
+ "should_training_stop": false
326
+ },
327
+ "attributes": {}
328
+ }
329
+ },
330
+ "total_flos": 0.0,
331
+ "train_batch_size": 1,
332
+ "trial_name": null,
333
+ "trial_params": null
334
+ }
gemma-9b-dpo/checkpoint-100/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74ee585106626f1196ffc9833586c5c11146fc384e5b33ed522cf45bab148032
3
+ size 6097
gemma-9b-dpo/checkpoint-150/README.md ADDED
@@ -0,0 +1,209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: google/gemma-2-9b-it
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - base_model:adapter:google/gemma-2-9b-it
7
+ - dpo
8
+ - lora
9
+ - transformers
10
+ - trl
11
+ ---
12
+
13
+ # Model Card for Model ID
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+
19
+ ## Model Details
20
+
21
+ ### Model Description
22
+
23
+ <!-- Provide a longer summary of what this model is. -->
24
+
25
+
26
+
27
+ - **Developed by:** [More Information Needed]
28
+ - **Funded by [optional]:** [More Information Needed]
29
+ - **Shared by [optional]:** [More Information Needed]
30
+ - **Model type:** [More Information Needed]
31
+ - **Language(s) (NLP):** [More Information Needed]
32
+ - **License:** [More Information Needed]
33
+ - **Finetuned from model [optional]:** [More Information Needed]
34
+
35
+ ### Model Sources [optional]
36
+
37
+ <!-- Provide the basic links for the model. -->
38
+
39
+ - **Repository:** [More Information Needed]
40
+ - **Paper [optional]:** [More Information Needed]
41
+ - **Demo [optional]:** [More Information Needed]
42
+
43
+ ## Uses
44
+
45
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
46
+
47
+ ### Direct Use
48
+
49
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
50
+
51
+ [More Information Needed]
52
+
53
+ ### Downstream Use [optional]
54
+
55
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
56
+
57
+ [More Information Needed]
58
+
59
+ ### Out-of-Scope Use
60
+
61
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
62
+
63
+ [More Information Needed]
64
+
65
+ ## Bias, Risks, and Limitations
66
+
67
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
68
+
69
+ [More Information Needed]
70
+
71
+ ### Recommendations
72
+
73
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
74
+
75
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
76
+
77
+ ## How to Get Started with the Model
78
+
79
+ Use the code below to get started with the model.
80
+
81
+ [More Information Needed]
82
+
83
+ ## Training Details
84
+
85
+ ### Training Data
86
+
87
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
88
+
89
+ [More Information Needed]
90
+
91
+ ### Training Procedure
92
+
93
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
94
+
95
+ #### Preprocessing [optional]
96
+
97
+ [More Information Needed]
98
+
99
+
100
+ #### Training Hyperparameters
101
+
102
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
103
+
104
+ #### Speeds, Sizes, Times [optional]
105
+
106
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
107
+
108
+ [More Information Needed]
109
+
110
+ ## Evaluation
111
+
112
+ <!-- This section describes the evaluation protocols and provides the results. -->
113
+
114
+ ### Testing Data, Factors & Metrics
115
+
116
+ #### Testing Data
117
+
118
+ <!-- This should link to a Dataset Card if possible. -->
119
+
120
+ [More Information Needed]
121
+
122
+ #### Factors
123
+
124
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
125
+
126
+ [More Information Needed]
127
+
128
+ #### Metrics
129
+
130
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
131
+
132
+ [More Information Needed]
133
+
134
+ ### Results
135
+
136
+ [More Information Needed]
137
+
138
+ #### Summary
139
+
140
+
141
+
142
+ ## Model Examination [optional]
143
+
144
+ <!-- Relevant interpretability work for the model goes here -->
145
+
146
+ [More Information Needed]
147
+
148
+ ## Environmental Impact
149
+
150
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
151
+
152
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
153
+
154
+ - **Hardware Type:** [More Information Needed]
155
+ - **Hours used:** [More Information Needed]
156
+ - **Cloud Provider:** [More Information Needed]
157
+ - **Compute Region:** [More Information Needed]
158
+ - **Carbon Emitted:** [More Information Needed]
159
+
160
+ ## Technical Specifications [optional]
161
+
162
+ ### Model Architecture and Objective
163
+
164
+ [More Information Needed]
165
+
166
+ ### Compute Infrastructure
167
+
168
+ [More Information Needed]
169
+
170
+ #### Hardware
171
+
172
+ [More Information Needed]
173
+
174
+ #### Software
175
+
176
+ [More Information Needed]
177
+
178
+ ## Citation [optional]
179
+
180
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
181
+
182
+ **BibTeX:**
183
+
184
+ [More Information Needed]
185
+
186
+ **APA:**
187
+
188
+ [More Information Needed]
189
+
190
+ ## Glossary [optional]
191
+
192
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
193
+
194
+ [More Information Needed]
195
+
196
+ ## More Information [optional]
197
+
198
+ [More Information Needed]
199
+
200
+ ## Model Card Authors [optional]
201
+
202
+ [More Information Needed]
203
+
204
+ ## Model Card Contact
205
+
206
+ [More Information Needed]
207
+ ### Framework versions
208
+
209
+ - PEFT 0.18.1
gemma-9b-dpo/checkpoint-150/adapter_config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "google/gemma-2-9b-it",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": false,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 32,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.05,
22
+ "megatron_config": null,
23
+ "megatron_core": "megatron.core",
24
+ "modules_to_save": null,
25
+ "peft_type": "LORA",
26
+ "peft_version": "0.18.1",
27
+ "qalora_group_size": 16,
28
+ "r": 16,
29
+ "rank_pattern": {},
30
+ "revision": null,
31
+ "target_modules": [
32
+ "down_proj",
33
+ "v_proj",
34
+ "k_proj",
35
+ "up_proj",
36
+ "q_proj",
37
+ "gate_proj",
38
+ "o_proj"
39
+ ],
40
+ "target_parameters": null,
41
+ "task_type": "CAUSAL_LM",
42
+ "trainable_token_indices": null,
43
+ "use_dora": false,
44
+ "use_qalora": false,
45
+ "use_rslora": false
46
+ }
gemma-9b-dpo/checkpoint-150/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fb4b6a23e87c3afa5d1d7677ba18415494ce40fd39710929e6f25cd9d556c70
3
+ size 216151256
gemma-9b-dpo/checkpoint-150/chat_template.jinja ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {{ bos_token }}{% if messages[0]['role'] == 'system' %}{{ raise_exception('System role not supported') }}{% endif %}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if (message['role'] == 'assistant') %}{% set role = 'model' %}{% else %}{% set role = message['role'] %}{% endif %}{{ '<start_of_turn>' + role + '
2
+ ' + message['content'] | trim + '<end_of_turn>
3
+ ' }}{% endfor %}{% if add_generation_prompt %}{{'<start_of_turn>model
4
+ '}}{% endif %}
gemma-9b-dpo/checkpoint-150/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:650dffa633e77bcfdec0baf6affbd38388fc15c92b1de262b46fc28615ccf125
3
+ size 110425877
gemma-9b-dpo/checkpoint-150/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de69a2834426ff9ef8199d077e00892579278af31d8969d77f98235b5cfc010a
3
+ size 14645
gemma-9b-dpo/checkpoint-150/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bddf86a69e8937c848fd7d73834147cc9d2b9b28c3b046c3f4e213af6efdce30
3
+ size 1465
gemma-9b-dpo/checkpoint-150/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:394ace002a144ac6ad5486387502f2d36f70c087310c3d907857240c76fcb36e
3
+ size 34362748
gemma-9b-dpo/checkpoint-150/tokenizer_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "<bos>",
4
+ "clean_up_tokenization_spaces": false,
5
+ "eos_token": "<eos>",
6
+ "extra_special_tokens": [
7
+ "<start_of_turn>",
8
+ "<end_of_turn>"
9
+ ],
10
+ "is_local": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 1000000000000000019884624838656,
13
+ "pad_token": "<pad>",
14
+ "sp_model_kwargs": {},
15
+ "spaces_between_special_tokens": false,
16
+ "tokenizer_class": "GemmaTokenizer",
17
+ "unk_token": "<unk>",
18
+ "use_default_system_prompt": false
19
+ }
gemma-9b-dpo/checkpoint-150/trainer_state.json ADDED
@@ -0,0 +1,484 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.2838709677419355,
6
+ "eval_steps": 500,
7
+ "global_step": 150,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.043010752688172046,
14
+ "grad_norm": 1.6562875509262085,
15
+ "learning_rate": 3.333333333333333e-07,
16
+ "logits/chosen": -5.94815731048584,
17
+ "logits/rejected": -5.856410503387451,
18
+ "logps/chosen": -302.07794189453125,
19
+ "logps/rejected": -283.5802001953125,
20
+ "loss": 0.6852486610412598,
21
+ "rewards/accuracies": 0.375,
22
+ "rewards/chosen": 0.01349079143255949,
23
+ "rewards/margins": 0.017757166177034378,
24
+ "rewards/rejected": -0.00426637614145875,
25
+ "step": 5
26
+ },
27
+ {
28
+ "epoch": 0.08602150537634409,
29
+ "grad_norm": 1.7723957300186157,
30
+ "learning_rate": 7.5e-07,
31
+ "logits/chosen": -5.585428714752197,
32
+ "logits/rejected": -5.577895641326904,
33
+ "logps/chosen": -362.89239501953125,
34
+ "logps/rejected": -280.8010559082031,
35
+ "loss": 0.6844330310821534,
36
+ "rewards/accuracies": 0.6499999761581421,
37
+ "rewards/chosen": 0.018745478242635727,
38
+ "rewards/margins": 0.018739622086286545,
39
+ "rewards/rejected": 5.85438692723983e-06,
40
+ "step": 10
41
+ },
42
+ {
43
+ "epoch": 0.12903225806451613,
44
+ "grad_norm": 1.6630758047103882,
45
+ "learning_rate": 1.1666666666666668e-06,
46
+ "logits/chosen": -5.916059494018555,
47
+ "logits/rejected": -5.797668933868408,
48
+ "logps/chosen": -307.4602966308594,
49
+ "logps/rejected": -316.79803466796875,
50
+ "loss": 0.7070020198822021,
51
+ "rewards/accuracies": 0.5249999761581421,
52
+ "rewards/chosen": -0.008706416003406048,
53
+ "rewards/margins": -0.02138950116932392,
54
+ "rewards/rejected": 0.012683087959885597,
55
+ "step": 15
56
+ },
57
+ {
58
+ "epoch": 0.17204301075268819,
59
+ "grad_norm": 5.225494861602783,
60
+ "learning_rate": 1.5833333333333333e-06,
61
+ "logits/chosen": -5.818185329437256,
62
+ "logits/rejected": -5.638014793395996,
63
+ "logps/chosen": -329.3303527832031,
64
+ "logps/rejected": -292.2645568847656,
65
+ "loss": 0.6929943561553955,
66
+ "rewards/accuracies": 0.4749999940395355,
67
+ "rewards/chosen": -0.044999174773693085,
68
+ "rewards/margins": 0.00216915225610137,
69
+ "rewards/rejected": -0.04716832935810089,
70
+ "step": 20
71
+ },
72
+ {
73
+ "epoch": 0.21505376344086022,
74
+ "grad_norm": 2.5229244232177734,
75
+ "learning_rate": 2e-06,
76
+ "logits/chosen": -5.8189592361450195,
77
+ "logits/rejected": -5.646462917327881,
78
+ "logps/chosen": -311.9922180175781,
79
+ "logps/rejected": -297.0098571777344,
80
+ "loss": 0.7002316951751709,
81
+ "rewards/accuracies": 0.4000000059604645,
82
+ "rewards/chosen": 0.0544702522456646,
83
+ "rewards/margins": -0.012132339179515839,
84
+ "rewards/rejected": 0.06660258769989014,
85
+ "step": 25
86
+ },
87
+ {
88
+ "epoch": 0.25806451612903225,
89
+ "grad_norm": 1.6524324417114258,
90
+ "learning_rate": 2.4166666666666667e-06,
91
+ "logits/chosen": -5.696784496307373,
92
+ "logits/rejected": -5.6943678855896,
93
+ "logps/chosen": -350.2938537597656,
94
+ "logps/rejected": -322.43182373046875,
95
+ "loss": 0.699475908279419,
96
+ "rewards/accuracies": 0.574999988079071,
97
+ "rewards/chosen": -0.0020792910363525152,
98
+ "rewards/margins": 0.02348965033888817,
99
+ "rewards/rejected": -0.025568943470716476,
100
+ "step": 30
101
+ },
102
+ {
103
+ "epoch": 0.3010752688172043,
104
+ "grad_norm": 9.075493812561035,
105
+ "learning_rate": 2.8333333333333335e-06,
106
+ "logits/chosen": -5.407717704772949,
107
+ "logits/rejected": -5.5642805099487305,
108
+ "logps/chosen": -391.3837585449219,
109
+ "logps/rejected": -333.9232482910156,
110
+ "loss": 0.6926415920257568,
111
+ "rewards/accuracies": 0.4749999940395355,
112
+ "rewards/chosen": -0.026233959943056107,
113
+ "rewards/margins": 0.00582819152623415,
114
+ "rewards/rejected": -0.03206215053796768,
115
+ "step": 35
116
+ },
117
+ {
118
+ "epoch": 0.34408602150537637,
119
+ "grad_norm": 2.6282100677490234,
120
+ "learning_rate": 2.9714285714285716e-06,
121
+ "logits/chosen": -5.649778366088867,
122
+ "logits/rejected": -5.512002944946289,
123
+ "logps/chosen": -338.31048583984375,
124
+ "logps/rejected": -279.06536865234375,
125
+ "loss": 0.6739111423492432,
126
+ "rewards/accuracies": 0.625,
127
+ "rewards/chosen": 0.04578382521867752,
128
+ "rewards/margins": 0.042618148028850555,
129
+ "rewards/rejected": 0.0031656839419156313,
130
+ "step": 40
131
+ },
132
+ {
133
+ "epoch": 0.3870967741935484,
134
+ "grad_norm": 16.96689796447754,
135
+ "learning_rate": 2.923809523809524e-06,
136
+ "logits/chosen": -5.518254280090332,
137
+ "logits/rejected": -5.577255725860596,
138
+ "logps/chosen": -374.29449462890625,
139
+ "logps/rejected": -366.4861145019531,
140
+ "loss": 0.7162120819091797,
141
+ "rewards/accuracies": 0.550000011920929,
142
+ "rewards/chosen": 0.03416462987661362,
143
+ "rewards/margins": 0.0036566159687936306,
144
+ "rewards/rejected": 0.030508000403642654,
145
+ "step": 45
146
+ },
147
+ {
148
+ "epoch": 0.43010752688172044,
149
+ "grad_norm": 1.447816252708435,
150
+ "learning_rate": 2.8761904761904764e-06,
151
+ "logits/chosen": -5.831571102142334,
152
+ "logits/rejected": -5.802765846252441,
153
+ "logps/chosen": -316.0965881347656,
154
+ "logps/rejected": -269.0494384765625,
155
+ "loss": 0.6838803291320801,
156
+ "rewards/accuracies": 0.550000011920929,
157
+ "rewards/chosen": -0.1265859305858612,
158
+ "rewards/margins": 0.026753634214401245,
159
+ "rewards/rejected": -0.15333956480026245,
160
+ "step": 50
161
+ },
162
+ {
163
+ "epoch": 0.4731182795698925,
164
+ "grad_norm": 2.174459457397461,
165
+ "learning_rate": 2.8285714285714288e-06,
166
+ "logits/chosen": -5.777364253997803,
167
+ "logits/rejected": -5.834509372711182,
168
+ "logps/chosen": -316.31390380859375,
169
+ "logps/rejected": -282.5376892089844,
170
+ "loss": 0.6832016944885254,
171
+ "rewards/accuracies": 0.675000011920929,
172
+ "rewards/chosen": -0.01756533980369568,
173
+ "rewards/margins": 0.02713041380047798,
174
+ "rewards/rejected": -0.04469575732946396,
175
+ "step": 55
176
+ },
177
+ {
178
+ "epoch": 0.5161290322580645,
179
+ "grad_norm": 2.2876009941101074,
180
+ "learning_rate": 2.780952380952381e-06,
181
+ "logits/chosen": -5.92340087890625,
182
+ "logits/rejected": -5.846226692199707,
183
+ "logps/chosen": -235.49667358398438,
184
+ "logps/rejected": -267.7764892578125,
185
+ "loss": 0.6656109809875488,
186
+ "rewards/accuracies": 0.6499999761581421,
187
+ "rewards/chosen": 0.021378064528107643,
188
+ "rewards/margins": 0.06001085042953491,
189
+ "rewards/rejected": -0.038632798939943314,
190
+ "step": 60
191
+ },
192
+ {
193
+ "epoch": 0.5591397849462365,
194
+ "grad_norm": 3.45564603805542,
195
+ "learning_rate": 2.733333333333333e-06,
196
+ "logits/chosen": -5.650258541107178,
197
+ "logits/rejected": -5.6721577644348145,
198
+ "logps/chosen": -381.40655517578125,
199
+ "logps/rejected": -311.68865966796875,
200
+ "loss": 0.7090614795684814,
201
+ "rewards/accuracies": 0.4749999940395355,
202
+ "rewards/chosen": 0.1167660504579544,
203
+ "rewards/margins": -0.0002812861348502338,
204
+ "rewards/rejected": 0.11704733222723007,
205
+ "step": 65
206
+ },
207
+ {
208
+ "epoch": 0.6021505376344086,
209
+ "grad_norm": 4.502689838409424,
210
+ "learning_rate": 2.685714285714286e-06,
211
+ "logits/chosen": -5.727735996246338,
212
+ "logits/rejected": -5.644078254699707,
213
+ "logps/chosen": -362.89886474609375,
214
+ "logps/rejected": -314.7126770019531,
215
+ "loss": 0.6596640110015869,
216
+ "rewards/accuracies": 0.6499999761581421,
217
+ "rewards/chosen": 0.07285688072443008,
218
+ "rewards/margins": 0.07270809262990952,
219
+ "rewards/rejected": 0.00014879256195854396,
220
+ "step": 70
221
+ },
222
+ {
223
+ "epoch": 0.6451612903225806,
224
+ "grad_norm": 2.2681005001068115,
225
+ "learning_rate": 2.638095238095238e-06,
226
+ "logits/chosen": -5.528594017028809,
227
+ "logits/rejected": -5.29849910736084,
228
+ "logps/chosen": -353.0726013183594,
229
+ "logps/rejected": -331.8568115234375,
230
+ "loss": 0.7387234687805175,
231
+ "rewards/accuracies": 0.574999988079071,
232
+ "rewards/chosen": 0.1685911864042282,
233
+ "rewards/margins": -0.04706493765115738,
234
+ "rewards/rejected": 0.2156561315059662,
235
+ "step": 75
236
+ },
237
+ {
238
+ "epoch": 0.6881720430107527,
239
+ "grad_norm": 3.0135886669158936,
240
+ "learning_rate": 2.5904761904761907e-06,
241
+ "logits/chosen": -5.8677144050598145,
242
+ "logits/rejected": -5.5384016036987305,
243
+ "logps/chosen": -326.17724609375,
244
+ "logps/rejected": -285.8974304199219,
245
+ "loss": 0.6047093391418457,
246
+ "rewards/accuracies": 0.925000011920929,
247
+ "rewards/chosen": 0.19864055514335632,
248
+ "rewards/margins": 0.19678126275539398,
249
+ "rewards/rejected": 0.0018593042623251677,
250
+ "step": 80
251
+ },
252
+ {
253
+ "epoch": 0.7311827956989247,
254
+ "grad_norm": 1.7684988975524902,
255
+ "learning_rate": 2.5428571428571427e-06,
256
+ "logits/chosen": -5.413943290710449,
257
+ "logits/rejected": -5.427316665649414,
258
+ "logps/chosen": -372.1170654296875,
259
+ "logps/rejected": -312.30279541015625,
260
+ "loss": 0.651799201965332,
261
+ "rewards/accuracies": 0.675000011920929,
262
+ "rewards/chosen": 0.179380863904953,
263
+ "rewards/margins": 0.09846550226211548,
264
+ "rewards/rejected": 0.08091535419225693,
265
+ "step": 85
266
+ },
267
+ {
268
+ "epoch": 0.7741935483870968,
269
+ "grad_norm": 2.4072341918945312,
270
+ "learning_rate": 2.4952380952380955e-06,
271
+ "logits/chosen": -5.4685492515563965,
272
+ "logits/rejected": -5.51275634765625,
273
+ "logps/chosen": -345.37188720703125,
274
+ "logps/rejected": -273.87261962890625,
275
+ "loss": 0.6362700462341309,
276
+ "rewards/accuracies": 0.6499999761581421,
277
+ "rewards/chosen": 0.33122482895851135,
278
+ "rewards/margins": 0.1338074505329132,
279
+ "rewards/rejected": 0.19741736352443695,
280
+ "step": 90
281
+ },
282
+ {
283
+ "epoch": 0.8172043010752689,
284
+ "grad_norm": 1.745698094367981,
285
+ "learning_rate": 2.4476190476190475e-06,
286
+ "logits/chosen": -5.6481757164001465,
287
+ "logits/rejected": -5.4260029792785645,
288
+ "logps/chosen": -337.11767578125,
289
+ "logps/rejected": -263.3458251953125,
290
+ "loss": 0.6222721099853515,
291
+ "rewards/accuracies": 0.824999988079071,
292
+ "rewards/chosen": 0.25297099351882935,
293
+ "rewards/margins": 0.1701255589723587,
294
+ "rewards/rejected": 0.08284540474414825,
295
+ "step": 95
296
+ },
297
+ {
298
+ "epoch": 0.8602150537634409,
299
+ "grad_norm": 1.4592756032943726,
300
+ "learning_rate": 2.4000000000000003e-06,
301
+ "logits/chosen": -5.402789115905762,
302
+ "logits/rejected": -5.280846118927002,
303
+ "logps/chosen": -304.55316162109375,
304
+ "logps/rejected": -314.95306396484375,
305
+ "loss": 0.6404934883117676,
306
+ "rewards/accuracies": 0.699999988079071,
307
+ "rewards/chosen": 0.29103565216064453,
308
+ "rewards/margins": 0.12328717857599258,
309
+ "rewards/rejected": 0.16774848103523254,
310
+ "step": 100
311
+ },
312
+ {
313
+ "epoch": 0.9032258064516129,
314
+ "grad_norm": 7.651180744171143,
315
+ "learning_rate": 2.3523809523809523e-06,
316
+ "logits/chosen": -5.333284854888916,
317
+ "logits/rejected": -5.209356307983398,
318
+ "logps/chosen": -385.76483154296875,
319
+ "logps/rejected": -317.0154724121094,
320
+ "loss": 0.5562876224517822,
321
+ "rewards/accuracies": 0.824999988079071,
322
+ "rewards/chosen": 0.5905097126960754,
323
+ "rewards/margins": 0.39645156264305115,
324
+ "rewards/rejected": 0.19405809044837952,
325
+ "step": 105
326
+ },
327
+ {
328
+ "epoch": 0.946236559139785,
329
+ "grad_norm": 2.2162137031555176,
330
+ "learning_rate": 2.304761904761905e-06,
331
+ "logits/chosen": -5.9164533615112305,
332
+ "logits/rejected": -5.660351276397705,
333
+ "logps/chosen": -362.57720947265625,
334
+ "logps/rejected": -328.386962890625,
335
+ "loss": 0.5672832489013672,
336
+ "rewards/accuracies": 0.824999988079071,
337
+ "rewards/chosen": 0.4752024710178375,
338
+ "rewards/margins": 0.3182791471481323,
339
+ "rewards/rejected": 0.15692326426506042,
340
+ "step": 110
341
+ },
342
+ {
343
+ "epoch": 0.989247311827957,
344
+ "grad_norm": 2.102505683898926,
345
+ "learning_rate": 2.257142857142857e-06,
346
+ "logits/chosen": -5.367539882659912,
347
+ "logits/rejected": -5.304169654846191,
348
+ "logps/chosen": -389.9139404296875,
349
+ "logps/rejected": -303.767822265625,
350
+ "loss": 0.563088321685791,
351
+ "rewards/accuracies": 0.800000011920929,
352
+ "rewards/chosen": 0.9107279777526855,
353
+ "rewards/margins": 0.7586480379104614,
354
+ "rewards/rejected": 0.15207989513874054,
355
+ "step": 115
356
+ },
357
+ {
358
+ "epoch": 1.0258064516129033,
359
+ "grad_norm": 2.9593698978424072,
360
+ "learning_rate": 2.20952380952381e-06,
361
+ "logits/chosen": -5.506618976593018,
362
+ "logits/rejected": -5.349832057952881,
363
+ "logps/chosen": -299.4593505859375,
364
+ "logps/rejected": -229.88784790039062,
365
+ "loss": 0.5665022850036621,
366
+ "rewards/accuracies": 0.8823529481887817,
367
+ "rewards/chosen": 0.4371771514415741,
368
+ "rewards/margins": 0.3876599073410034,
369
+ "rewards/rejected": 0.04951724037528038,
370
+ "step": 120
371
+ },
372
+ {
373
+ "epoch": 1.0688172043010753,
374
+ "grad_norm": 2.5988082885742188,
375
+ "learning_rate": 2.161904761904762e-06,
376
+ "logits/chosen": -5.542575836181641,
377
+ "logits/rejected": -5.428658485412598,
378
+ "logps/chosen": -353.8345642089844,
379
+ "logps/rejected": -312.9586486816406,
380
+ "loss": 0.5137276172637939,
381
+ "rewards/accuracies": 0.925000011920929,
382
+ "rewards/chosen": 0.6325557827949524,
383
+ "rewards/margins": 0.44065460562705994,
384
+ "rewards/rejected": 0.19190113246440887,
385
+ "step": 125
386
+ },
387
+ {
388
+ "epoch": 1.1118279569892473,
389
+ "grad_norm": 1.6948280334472656,
390
+ "learning_rate": 2.1142857142857147e-06,
391
+ "logits/chosen": -5.847611427307129,
392
+ "logits/rejected": -5.498036861419678,
393
+ "logps/chosen": -300.35577392578125,
394
+ "logps/rejected": -311.05126953125,
395
+ "loss": 0.5773736953735351,
396
+ "rewards/accuracies": 0.875,
397
+ "rewards/chosen": 0.6476074457168579,
398
+ "rewards/margins": 0.37814000248908997,
399
+ "rewards/rejected": 0.2694675028324127,
400
+ "step": 130
401
+ },
402
+ {
403
+ "epoch": 1.1548387096774193,
404
+ "grad_norm": 1.372768759727478,
405
+ "learning_rate": 2.0666666666666666e-06,
406
+ "logits/chosen": -5.619741916656494,
407
+ "logits/rejected": -5.618372917175293,
408
+ "logps/chosen": -367.06732177734375,
409
+ "logps/rejected": -301.2478332519531,
410
+ "loss": 0.46464052200317385,
411
+ "rewards/accuracies": 0.949999988079071,
412
+ "rewards/chosen": 0.8353813290596008,
413
+ "rewards/margins": 0.6073431968688965,
414
+ "rewards/rejected": 0.22803807258605957,
415
+ "step": 135
416
+ },
417
+ {
418
+ "epoch": 1.1978494623655913,
419
+ "grad_norm": 2.498854637145996,
420
+ "learning_rate": 2.0190476190476195e-06,
421
+ "logits/chosen": -5.195496559143066,
422
+ "logits/rejected": -4.919422149658203,
423
+ "logps/chosen": -333.09979248046875,
424
+ "logps/rejected": -317.78167724609375,
425
+ "loss": 0.5423533916473389,
426
+ "rewards/accuracies": 0.8500000238418579,
427
+ "rewards/chosen": 0.46827277541160583,
428
+ "rewards/margins": 0.42974653840065,
429
+ "rewards/rejected": 0.0385262668132782,
430
+ "step": 140
431
+ },
432
+ {
433
+ "epoch": 1.2408602150537635,
434
+ "grad_norm": 1.451978087425232,
435
+ "learning_rate": 1.9714285714285714e-06,
436
+ "logits/chosen": -5.425684452056885,
437
+ "logits/rejected": -5.322096824645996,
438
+ "logps/chosen": -314.4623107910156,
439
+ "logps/rejected": -271.85479736328125,
440
+ "loss": 0.4632129192352295,
441
+ "rewards/accuracies": 0.949999988079071,
442
+ "rewards/chosen": 0.9266977310180664,
443
+ "rewards/margins": 0.672536313533783,
444
+ "rewards/rejected": 0.25416144728660583,
445
+ "step": 145
446
+ },
447
+ {
448
+ "epoch": 1.2838709677419355,
449
+ "grad_norm": 2.0581016540527344,
450
+ "learning_rate": 1.923809523809524e-06,
451
+ "logits/chosen": -4.949021339416504,
452
+ "logits/rejected": -5.033829212188721,
453
+ "logps/chosen": -356.3919982910156,
454
+ "logps/rejected": -330.82720947265625,
455
+ "loss": 0.4621857166290283,
456
+ "rewards/accuracies": 0.8999999761581421,
457
+ "rewards/chosen": 0.7969595789909363,
458
+ "rewards/margins": 0.6875919103622437,
459
+ "rewards/rejected": 0.10936765372753143,
460
+ "step": 150
461
+ }
462
+ ],
463
+ "logging_steps": 5,
464
+ "max_steps": 351,
465
+ "num_input_tokens_seen": 0,
466
+ "num_train_epochs": 3,
467
+ "save_steps": 50,
468
+ "stateful_callbacks": {
469
+ "TrainerControl": {
470
+ "args": {
471
+ "should_epoch_stop": false,
472
+ "should_evaluate": false,
473
+ "should_log": false,
474
+ "should_save": true,
475
+ "should_training_stop": false
476
+ },
477
+ "attributes": {}
478
+ }
479
+ },
480
+ "total_flos": 0.0,
481
+ "train_batch_size": 1,
482
+ "trial_name": null,
483
+ "trial_params": null
484
+ }
gemma-9b-dpo/checkpoint-150/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74ee585106626f1196ffc9833586c5c11146fc384e5b33ed522cf45bab148032
3
+ size 6097
gemma-9b-dpo/checkpoint-200/README.md ADDED
@@ -0,0 +1,209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: google/gemma-2-9b-it
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - base_model:adapter:google/gemma-2-9b-it
7
+ - dpo
8
+ - lora
9
+ - transformers
10
+ - trl
11
+ ---
12
+
13
+ # Model Card for Model ID
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+
19
+ ## Model Details
20
+
21
+ ### Model Description
22
+
23
+ <!-- Provide a longer summary of what this model is. -->
24
+
25
+
26
+
27
+ - **Developed by:** [More Information Needed]
28
+ - **Funded by [optional]:** [More Information Needed]
29
+ - **Shared by [optional]:** [More Information Needed]
30
+ - **Model type:** [More Information Needed]
31
+ - **Language(s) (NLP):** [More Information Needed]
32
+ - **License:** [More Information Needed]
33
+ - **Finetuned from model [optional]:** [More Information Needed]
34
+
35
+ ### Model Sources [optional]
36
+
37
+ <!-- Provide the basic links for the model. -->
38
+
39
+ - **Repository:** [More Information Needed]
40
+ - **Paper [optional]:** [More Information Needed]
41
+ - **Demo [optional]:** [More Information Needed]
42
+
43
+ ## Uses
44
+
45
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
46
+
47
+ ### Direct Use
48
+
49
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
50
+
51
+ [More Information Needed]
52
+
53
+ ### Downstream Use [optional]
54
+
55
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
56
+
57
+ [More Information Needed]
58
+
59
+ ### Out-of-Scope Use
60
+
61
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
62
+
63
+ [More Information Needed]
64
+
65
+ ## Bias, Risks, and Limitations
66
+
67
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
68
+
69
+ [More Information Needed]
70
+
71
+ ### Recommendations
72
+
73
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
74
+
75
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
76
+
77
+ ## How to Get Started with the Model
78
+
79
+ Use the code below to get started with the model.
80
+
81
+ [More Information Needed]
82
+
83
+ ## Training Details
84
+
85
+ ### Training Data
86
+
87
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
88
+
89
+ [More Information Needed]
90
+
91
+ ### Training Procedure
92
+
93
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
94
+
95
+ #### Preprocessing [optional]
96
+
97
+ [More Information Needed]
98
+
99
+
100
+ #### Training Hyperparameters
101
+
102
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
103
+
104
+ #### Speeds, Sizes, Times [optional]
105
+
106
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
107
+
108
+ [More Information Needed]
109
+
110
+ ## Evaluation
111
+
112
+ <!-- This section describes the evaluation protocols and provides the results. -->
113
+
114
+ ### Testing Data, Factors & Metrics
115
+
116
+ #### Testing Data
117
+
118
+ <!-- This should link to a Dataset Card if possible. -->
119
+
120
+ [More Information Needed]
121
+
122
+ #### Factors
123
+
124
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
125
+
126
+ [More Information Needed]
127
+
128
+ #### Metrics
129
+
130
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
131
+
132
+ [More Information Needed]
133
+
134
+ ### Results
135
+
136
+ [More Information Needed]
137
+
138
+ #### Summary
139
+
140
+
141
+
142
+ ## Model Examination [optional]
143
+
144
+ <!-- Relevant interpretability work for the model goes here -->
145
+
146
+ [More Information Needed]
147
+
148
+ ## Environmental Impact
149
+
150
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
151
+
152
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
153
+
154
+ - **Hardware Type:** [More Information Needed]
155
+ - **Hours used:** [More Information Needed]
156
+ - **Cloud Provider:** [More Information Needed]
157
+ - **Compute Region:** [More Information Needed]
158
+ - **Carbon Emitted:** [More Information Needed]
159
+
160
+ ## Technical Specifications [optional]
161
+
162
+ ### Model Architecture and Objective
163
+
164
+ [More Information Needed]
165
+
166
+ ### Compute Infrastructure
167
+
168
+ [More Information Needed]
169
+
170
+ #### Hardware
171
+
172
+ [More Information Needed]
173
+
174
+ #### Software
175
+
176
+ [More Information Needed]
177
+
178
+ ## Citation [optional]
179
+
180
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
181
+
182
+ **BibTeX:**
183
+
184
+ [More Information Needed]
185
+
186
+ **APA:**
187
+
188
+ [More Information Needed]
189
+
190
+ ## Glossary [optional]
191
+
192
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
193
+
194
+ [More Information Needed]
195
+
196
+ ## More Information [optional]
197
+
198
+ [More Information Needed]
199
+
200
+ ## Model Card Authors [optional]
201
+
202
+ [More Information Needed]
203
+
204
+ ## Model Card Contact
205
+
206
+ [More Information Needed]
207
+ ### Framework versions
208
+
209
+ - PEFT 0.18.1
gemma-9b-dpo/checkpoint-200/adapter_config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "google/gemma-2-9b-it",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": false,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 32,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.05,
22
+ "megatron_config": null,
23
+ "megatron_core": "megatron.core",
24
+ "modules_to_save": null,
25
+ "peft_type": "LORA",
26
+ "peft_version": "0.18.1",
27
+ "qalora_group_size": 16,
28
+ "r": 16,
29
+ "rank_pattern": {},
30
+ "revision": null,
31
+ "target_modules": [
32
+ "down_proj",
33
+ "v_proj",
34
+ "k_proj",
35
+ "up_proj",
36
+ "q_proj",
37
+ "gate_proj",
38
+ "o_proj"
39
+ ],
40
+ "target_parameters": null,
41
+ "task_type": "CAUSAL_LM",
42
+ "trainable_token_indices": null,
43
+ "use_dora": false,
44
+ "use_qalora": false,
45
+ "use_rslora": false
46
+ }
gemma-9b-dpo/checkpoint-200/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:148d36290eaf7e7ff0d36aa2e69413dd96652ab1ec0ba672aa91c7703cea564e
3
+ size 216151256
gemma-9b-dpo/checkpoint-200/chat_template.jinja ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {{ bos_token }}{% if messages[0]['role'] == 'system' %}{{ raise_exception('System role not supported') }}{% endif %}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if (message['role'] == 'assistant') %}{% set role = 'model' %}{% else %}{% set role = message['role'] %}{% endif %}{{ '<start_of_turn>' + role + '
2
+ ' + message['content'] | trim + '<end_of_turn>
3
+ ' }}{% endfor %}{% if add_generation_prompt %}{{'<start_of_turn>model
4
+ '}}{% endif %}
gemma-9b-dpo/checkpoint-200/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30ca54e740d552dff3c08af09de59aa65078627e4eaa324f024b288ed4f01d76
3
+ size 110425877
gemma-9b-dpo/checkpoint-200/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de69a2834426ff9ef8199d077e00892579278af31d8969d77f98235b5cfc010a
3
+ size 14645
gemma-9b-dpo/checkpoint-200/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5591e056f6f3a8b46f89ab0e15a297b8ef9e5d149ccdfe0eb7df998b20025b07
3
+ size 1465
gemma-9b-dpo/checkpoint-200/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:394ace002a144ac6ad5486387502f2d36f70c087310c3d907857240c76fcb36e
3
+ size 34362748
gemma-9b-dpo/checkpoint-200/tokenizer_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "<bos>",
4
+ "clean_up_tokenization_spaces": false,
5
+ "eos_token": "<eos>",
6
+ "extra_special_tokens": [
7
+ "<start_of_turn>",
8
+ "<end_of_turn>"
9
+ ],
10
+ "is_local": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 1000000000000000019884624838656,
13
+ "pad_token": "<pad>",
14
+ "sp_model_kwargs": {},
15
+ "spaces_between_special_tokens": false,
16
+ "tokenizer_class": "GemmaTokenizer",
17
+ "unk_token": "<unk>",
18
+ "use_default_system_prompt": false
19
+ }
gemma-9b-dpo/checkpoint-200/trainer_state.json ADDED
@@ -0,0 +1,634 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.7139784946236558,
6
+ "eval_steps": 500,
7
+ "global_step": 200,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.043010752688172046,
14
+ "grad_norm": 1.6562875509262085,
15
+ "learning_rate": 3.333333333333333e-07,
16
+ "logits/chosen": -5.94815731048584,
17
+ "logits/rejected": -5.856410503387451,
18
+ "logps/chosen": -302.07794189453125,
19
+ "logps/rejected": -283.5802001953125,
20
+ "loss": 0.6852486610412598,
21
+ "rewards/accuracies": 0.375,
22
+ "rewards/chosen": 0.01349079143255949,
23
+ "rewards/margins": 0.017757166177034378,
24
+ "rewards/rejected": -0.00426637614145875,
25
+ "step": 5
26
+ },
27
+ {
28
+ "epoch": 0.08602150537634409,
29
+ "grad_norm": 1.7723957300186157,
30
+ "learning_rate": 7.5e-07,
31
+ "logits/chosen": -5.585428714752197,
32
+ "logits/rejected": -5.577895641326904,
33
+ "logps/chosen": -362.89239501953125,
34
+ "logps/rejected": -280.8010559082031,
35
+ "loss": 0.6844330310821534,
36
+ "rewards/accuracies": 0.6499999761581421,
37
+ "rewards/chosen": 0.018745478242635727,
38
+ "rewards/margins": 0.018739622086286545,
39
+ "rewards/rejected": 5.85438692723983e-06,
40
+ "step": 10
41
+ },
42
+ {
43
+ "epoch": 0.12903225806451613,
44
+ "grad_norm": 1.6630758047103882,
45
+ "learning_rate": 1.1666666666666668e-06,
46
+ "logits/chosen": -5.916059494018555,
47
+ "logits/rejected": -5.797668933868408,
48
+ "logps/chosen": -307.4602966308594,
49
+ "logps/rejected": -316.79803466796875,
50
+ "loss": 0.7070020198822021,
51
+ "rewards/accuracies": 0.5249999761581421,
52
+ "rewards/chosen": -0.008706416003406048,
53
+ "rewards/margins": -0.02138950116932392,
54
+ "rewards/rejected": 0.012683087959885597,
55
+ "step": 15
56
+ },
57
+ {
58
+ "epoch": 0.17204301075268819,
59
+ "grad_norm": 5.225494861602783,
60
+ "learning_rate": 1.5833333333333333e-06,
61
+ "logits/chosen": -5.818185329437256,
62
+ "logits/rejected": -5.638014793395996,
63
+ "logps/chosen": -329.3303527832031,
64
+ "logps/rejected": -292.2645568847656,
65
+ "loss": 0.6929943561553955,
66
+ "rewards/accuracies": 0.4749999940395355,
67
+ "rewards/chosen": -0.044999174773693085,
68
+ "rewards/margins": 0.00216915225610137,
69
+ "rewards/rejected": -0.04716832935810089,
70
+ "step": 20
71
+ },
72
+ {
73
+ "epoch": 0.21505376344086022,
74
+ "grad_norm": 2.5229244232177734,
75
+ "learning_rate": 2e-06,
76
+ "logits/chosen": -5.8189592361450195,
77
+ "logits/rejected": -5.646462917327881,
78
+ "logps/chosen": -311.9922180175781,
79
+ "logps/rejected": -297.0098571777344,
80
+ "loss": 0.7002316951751709,
81
+ "rewards/accuracies": 0.4000000059604645,
82
+ "rewards/chosen": 0.0544702522456646,
83
+ "rewards/margins": -0.012132339179515839,
84
+ "rewards/rejected": 0.06660258769989014,
85
+ "step": 25
86
+ },
87
+ {
88
+ "epoch": 0.25806451612903225,
89
+ "grad_norm": 1.6524324417114258,
90
+ "learning_rate": 2.4166666666666667e-06,
91
+ "logits/chosen": -5.696784496307373,
92
+ "logits/rejected": -5.6943678855896,
93
+ "logps/chosen": -350.2938537597656,
94
+ "logps/rejected": -322.43182373046875,
95
+ "loss": 0.699475908279419,
96
+ "rewards/accuracies": 0.574999988079071,
97
+ "rewards/chosen": -0.0020792910363525152,
98
+ "rewards/margins": 0.02348965033888817,
99
+ "rewards/rejected": -0.025568943470716476,
100
+ "step": 30
101
+ },
102
+ {
103
+ "epoch": 0.3010752688172043,
104
+ "grad_norm": 9.075493812561035,
105
+ "learning_rate": 2.8333333333333335e-06,
106
+ "logits/chosen": -5.407717704772949,
107
+ "logits/rejected": -5.5642805099487305,
108
+ "logps/chosen": -391.3837585449219,
109
+ "logps/rejected": -333.9232482910156,
110
+ "loss": 0.6926415920257568,
111
+ "rewards/accuracies": 0.4749999940395355,
112
+ "rewards/chosen": -0.026233959943056107,
113
+ "rewards/margins": 0.00582819152623415,
114
+ "rewards/rejected": -0.03206215053796768,
115
+ "step": 35
116
+ },
117
+ {
118
+ "epoch": 0.34408602150537637,
119
+ "grad_norm": 2.6282100677490234,
120
+ "learning_rate": 2.9714285714285716e-06,
121
+ "logits/chosen": -5.649778366088867,
122
+ "logits/rejected": -5.512002944946289,
123
+ "logps/chosen": -338.31048583984375,
124
+ "logps/rejected": -279.06536865234375,
125
+ "loss": 0.6739111423492432,
126
+ "rewards/accuracies": 0.625,
127
+ "rewards/chosen": 0.04578382521867752,
128
+ "rewards/margins": 0.042618148028850555,
129
+ "rewards/rejected": 0.0031656839419156313,
130
+ "step": 40
131
+ },
132
+ {
133
+ "epoch": 0.3870967741935484,
134
+ "grad_norm": 16.96689796447754,
135
+ "learning_rate": 2.923809523809524e-06,
136
+ "logits/chosen": -5.518254280090332,
137
+ "logits/rejected": -5.577255725860596,
138
+ "logps/chosen": -374.29449462890625,
139
+ "logps/rejected": -366.4861145019531,
140
+ "loss": 0.7162120819091797,
141
+ "rewards/accuracies": 0.550000011920929,
142
+ "rewards/chosen": 0.03416462987661362,
143
+ "rewards/margins": 0.0036566159687936306,
144
+ "rewards/rejected": 0.030508000403642654,
145
+ "step": 45
146
+ },
147
+ {
148
+ "epoch": 0.43010752688172044,
149
+ "grad_norm": 1.447816252708435,
150
+ "learning_rate": 2.8761904761904764e-06,
151
+ "logits/chosen": -5.831571102142334,
152
+ "logits/rejected": -5.802765846252441,
153
+ "logps/chosen": -316.0965881347656,
154
+ "logps/rejected": -269.0494384765625,
155
+ "loss": 0.6838803291320801,
156
+ "rewards/accuracies": 0.550000011920929,
157
+ "rewards/chosen": -0.1265859305858612,
158
+ "rewards/margins": 0.026753634214401245,
159
+ "rewards/rejected": -0.15333956480026245,
160
+ "step": 50
161
+ },
162
+ {
163
+ "epoch": 0.4731182795698925,
164
+ "grad_norm": 2.174459457397461,
165
+ "learning_rate": 2.8285714285714288e-06,
166
+ "logits/chosen": -5.777364253997803,
167
+ "logits/rejected": -5.834509372711182,
168
+ "logps/chosen": -316.31390380859375,
169
+ "logps/rejected": -282.5376892089844,
170
+ "loss": 0.6832016944885254,
171
+ "rewards/accuracies": 0.675000011920929,
172
+ "rewards/chosen": -0.01756533980369568,
173
+ "rewards/margins": 0.02713041380047798,
174
+ "rewards/rejected": -0.04469575732946396,
175
+ "step": 55
176
+ },
177
+ {
178
+ "epoch": 0.5161290322580645,
179
+ "grad_norm": 2.2876009941101074,
180
+ "learning_rate": 2.780952380952381e-06,
181
+ "logits/chosen": -5.92340087890625,
182
+ "logits/rejected": -5.846226692199707,
183
+ "logps/chosen": -235.49667358398438,
184
+ "logps/rejected": -267.7764892578125,
185
+ "loss": 0.6656109809875488,
186
+ "rewards/accuracies": 0.6499999761581421,
187
+ "rewards/chosen": 0.021378064528107643,
188
+ "rewards/margins": 0.06001085042953491,
189
+ "rewards/rejected": -0.038632798939943314,
190
+ "step": 60
191
+ },
192
+ {
193
+ "epoch": 0.5591397849462365,
194
+ "grad_norm": 3.45564603805542,
195
+ "learning_rate": 2.733333333333333e-06,
196
+ "logits/chosen": -5.650258541107178,
197
+ "logits/rejected": -5.6721577644348145,
198
+ "logps/chosen": -381.40655517578125,
199
+ "logps/rejected": -311.68865966796875,
200
+ "loss": 0.7090614795684814,
201
+ "rewards/accuracies": 0.4749999940395355,
202
+ "rewards/chosen": 0.1167660504579544,
203
+ "rewards/margins": -0.0002812861348502338,
204
+ "rewards/rejected": 0.11704733222723007,
205
+ "step": 65
206
+ },
207
+ {
208
+ "epoch": 0.6021505376344086,
209
+ "grad_norm": 4.502689838409424,
210
+ "learning_rate": 2.685714285714286e-06,
211
+ "logits/chosen": -5.727735996246338,
212
+ "logits/rejected": -5.644078254699707,
213
+ "logps/chosen": -362.89886474609375,
214
+ "logps/rejected": -314.7126770019531,
215
+ "loss": 0.6596640110015869,
216
+ "rewards/accuracies": 0.6499999761581421,
217
+ "rewards/chosen": 0.07285688072443008,
218
+ "rewards/margins": 0.07270809262990952,
219
+ "rewards/rejected": 0.00014879256195854396,
220
+ "step": 70
221
+ },
222
+ {
223
+ "epoch": 0.6451612903225806,
224
+ "grad_norm": 2.2681005001068115,
225
+ "learning_rate": 2.638095238095238e-06,
226
+ "logits/chosen": -5.528594017028809,
227
+ "logits/rejected": -5.29849910736084,
228
+ "logps/chosen": -353.0726013183594,
229
+ "logps/rejected": -331.8568115234375,
230
+ "loss": 0.7387234687805175,
231
+ "rewards/accuracies": 0.574999988079071,
232
+ "rewards/chosen": 0.1685911864042282,
233
+ "rewards/margins": -0.04706493765115738,
234
+ "rewards/rejected": 0.2156561315059662,
235
+ "step": 75
236
+ },
237
+ {
238
+ "epoch": 0.6881720430107527,
239
+ "grad_norm": 3.0135886669158936,
240
+ "learning_rate": 2.5904761904761907e-06,
241
+ "logits/chosen": -5.8677144050598145,
242
+ "logits/rejected": -5.5384016036987305,
243
+ "logps/chosen": -326.17724609375,
244
+ "logps/rejected": -285.8974304199219,
245
+ "loss": 0.6047093391418457,
246
+ "rewards/accuracies": 0.925000011920929,
247
+ "rewards/chosen": 0.19864055514335632,
248
+ "rewards/margins": 0.19678126275539398,
249
+ "rewards/rejected": 0.0018593042623251677,
250
+ "step": 80
251
+ },
252
+ {
253
+ "epoch": 0.7311827956989247,
254
+ "grad_norm": 1.7684988975524902,
255
+ "learning_rate": 2.5428571428571427e-06,
256
+ "logits/chosen": -5.413943290710449,
257
+ "logits/rejected": -5.427316665649414,
258
+ "logps/chosen": -372.1170654296875,
259
+ "logps/rejected": -312.30279541015625,
260
+ "loss": 0.651799201965332,
261
+ "rewards/accuracies": 0.675000011920929,
262
+ "rewards/chosen": 0.179380863904953,
263
+ "rewards/margins": 0.09846550226211548,
264
+ "rewards/rejected": 0.08091535419225693,
265
+ "step": 85
266
+ },
267
+ {
268
+ "epoch": 0.7741935483870968,
269
+ "grad_norm": 2.4072341918945312,
270
+ "learning_rate": 2.4952380952380955e-06,
271
+ "logits/chosen": -5.4685492515563965,
272
+ "logits/rejected": -5.51275634765625,
273
+ "logps/chosen": -345.37188720703125,
274
+ "logps/rejected": -273.87261962890625,
275
+ "loss": 0.6362700462341309,
276
+ "rewards/accuracies": 0.6499999761581421,
277
+ "rewards/chosen": 0.33122482895851135,
278
+ "rewards/margins": 0.1338074505329132,
279
+ "rewards/rejected": 0.19741736352443695,
280
+ "step": 90
281
+ },
282
+ {
283
+ "epoch": 0.8172043010752689,
284
+ "grad_norm": 1.745698094367981,
285
+ "learning_rate": 2.4476190476190475e-06,
286
+ "logits/chosen": -5.6481757164001465,
287
+ "logits/rejected": -5.4260029792785645,
288
+ "logps/chosen": -337.11767578125,
289
+ "logps/rejected": -263.3458251953125,
290
+ "loss": 0.6222721099853515,
291
+ "rewards/accuracies": 0.824999988079071,
292
+ "rewards/chosen": 0.25297099351882935,
293
+ "rewards/margins": 0.1701255589723587,
294
+ "rewards/rejected": 0.08284540474414825,
295
+ "step": 95
296
+ },
297
+ {
298
+ "epoch": 0.8602150537634409,
299
+ "grad_norm": 1.4592756032943726,
300
+ "learning_rate": 2.4000000000000003e-06,
301
+ "logits/chosen": -5.402789115905762,
302
+ "logits/rejected": -5.280846118927002,
303
+ "logps/chosen": -304.55316162109375,
304
+ "logps/rejected": -314.95306396484375,
305
+ "loss": 0.6404934883117676,
306
+ "rewards/accuracies": 0.699999988079071,
307
+ "rewards/chosen": 0.29103565216064453,
308
+ "rewards/margins": 0.12328717857599258,
309
+ "rewards/rejected": 0.16774848103523254,
310
+ "step": 100
311
+ },
312
+ {
313
+ "epoch": 0.9032258064516129,
314
+ "grad_norm": 7.651180744171143,
315
+ "learning_rate": 2.3523809523809523e-06,
316
+ "logits/chosen": -5.333284854888916,
317
+ "logits/rejected": -5.209356307983398,
318
+ "logps/chosen": -385.76483154296875,
319
+ "logps/rejected": -317.0154724121094,
320
+ "loss": 0.5562876224517822,
321
+ "rewards/accuracies": 0.824999988079071,
322
+ "rewards/chosen": 0.5905097126960754,
323
+ "rewards/margins": 0.39645156264305115,
324
+ "rewards/rejected": 0.19405809044837952,
325
+ "step": 105
326
+ },
327
+ {
328
+ "epoch": 0.946236559139785,
329
+ "grad_norm": 2.2162137031555176,
330
+ "learning_rate": 2.304761904761905e-06,
331
+ "logits/chosen": -5.9164533615112305,
332
+ "logits/rejected": -5.660351276397705,
333
+ "logps/chosen": -362.57720947265625,
334
+ "logps/rejected": -328.386962890625,
335
+ "loss": 0.5672832489013672,
336
+ "rewards/accuracies": 0.824999988079071,
337
+ "rewards/chosen": 0.4752024710178375,
338
+ "rewards/margins": 0.3182791471481323,
339
+ "rewards/rejected": 0.15692326426506042,
340
+ "step": 110
341
+ },
342
+ {
343
+ "epoch": 0.989247311827957,
344
+ "grad_norm": 2.102505683898926,
345
+ "learning_rate": 2.257142857142857e-06,
346
+ "logits/chosen": -5.367539882659912,
347
+ "logits/rejected": -5.304169654846191,
348
+ "logps/chosen": -389.9139404296875,
349
+ "logps/rejected": -303.767822265625,
350
+ "loss": 0.563088321685791,
351
+ "rewards/accuracies": 0.800000011920929,
352
+ "rewards/chosen": 0.9107279777526855,
353
+ "rewards/margins": 0.7586480379104614,
354
+ "rewards/rejected": 0.15207989513874054,
355
+ "step": 115
356
+ },
357
+ {
358
+ "epoch": 1.0258064516129033,
359
+ "grad_norm": 2.9593698978424072,
360
+ "learning_rate": 2.20952380952381e-06,
361
+ "logits/chosen": -5.506618976593018,
362
+ "logits/rejected": -5.349832057952881,
363
+ "logps/chosen": -299.4593505859375,
364
+ "logps/rejected": -229.88784790039062,
365
+ "loss": 0.5665022850036621,
366
+ "rewards/accuracies": 0.8823529481887817,
367
+ "rewards/chosen": 0.4371771514415741,
368
+ "rewards/margins": 0.3876599073410034,
369
+ "rewards/rejected": 0.04951724037528038,
370
+ "step": 120
371
+ },
372
+ {
373
+ "epoch": 1.0688172043010753,
374
+ "grad_norm": 2.5988082885742188,
375
+ "learning_rate": 2.161904761904762e-06,
376
+ "logits/chosen": -5.542575836181641,
377
+ "logits/rejected": -5.428658485412598,
378
+ "logps/chosen": -353.8345642089844,
379
+ "logps/rejected": -312.9586486816406,
380
+ "loss": 0.5137276172637939,
381
+ "rewards/accuracies": 0.925000011920929,
382
+ "rewards/chosen": 0.6325557827949524,
383
+ "rewards/margins": 0.44065460562705994,
384
+ "rewards/rejected": 0.19190113246440887,
385
+ "step": 125
386
+ },
387
+ {
388
+ "epoch": 1.1118279569892473,
389
+ "grad_norm": 1.6948280334472656,
390
+ "learning_rate": 2.1142857142857147e-06,
391
+ "logits/chosen": -5.847611427307129,
392
+ "logits/rejected": -5.498036861419678,
393
+ "logps/chosen": -300.35577392578125,
394
+ "logps/rejected": -311.05126953125,
395
+ "loss": 0.5773736953735351,
396
+ "rewards/accuracies": 0.875,
397
+ "rewards/chosen": 0.6476074457168579,
398
+ "rewards/margins": 0.37814000248908997,
399
+ "rewards/rejected": 0.2694675028324127,
400
+ "step": 130
401
+ },
402
+ {
403
+ "epoch": 1.1548387096774193,
404
+ "grad_norm": 1.372768759727478,
405
+ "learning_rate": 2.0666666666666666e-06,
406
+ "logits/chosen": -5.619741916656494,
407
+ "logits/rejected": -5.618372917175293,
408
+ "logps/chosen": -367.06732177734375,
409
+ "logps/rejected": -301.2478332519531,
410
+ "loss": 0.46464052200317385,
411
+ "rewards/accuracies": 0.949999988079071,
412
+ "rewards/chosen": 0.8353813290596008,
413
+ "rewards/margins": 0.6073431968688965,
414
+ "rewards/rejected": 0.22803807258605957,
415
+ "step": 135
416
+ },
417
+ {
418
+ "epoch": 1.1978494623655913,
419
+ "grad_norm": 2.498854637145996,
420
+ "learning_rate": 2.0190476190476195e-06,
421
+ "logits/chosen": -5.195496559143066,
422
+ "logits/rejected": -4.919422149658203,
423
+ "logps/chosen": -333.09979248046875,
424
+ "logps/rejected": -317.78167724609375,
425
+ "loss": 0.5423533916473389,
426
+ "rewards/accuracies": 0.8500000238418579,
427
+ "rewards/chosen": 0.46827277541160583,
428
+ "rewards/margins": 0.42974653840065,
429
+ "rewards/rejected": 0.0385262668132782,
430
+ "step": 140
431
+ },
432
+ {
433
+ "epoch": 1.2408602150537635,
434
+ "grad_norm": 1.451978087425232,
435
+ "learning_rate": 1.9714285714285714e-06,
436
+ "logits/chosen": -5.425684452056885,
437
+ "logits/rejected": -5.322096824645996,
438
+ "logps/chosen": -314.4623107910156,
439
+ "logps/rejected": -271.85479736328125,
440
+ "loss": 0.4632129192352295,
441
+ "rewards/accuracies": 0.949999988079071,
442
+ "rewards/chosen": 0.9266977310180664,
443
+ "rewards/margins": 0.672536313533783,
444
+ "rewards/rejected": 0.25416144728660583,
445
+ "step": 145
446
+ },
447
+ {
448
+ "epoch": 1.2838709677419355,
449
+ "grad_norm": 2.0581016540527344,
450
+ "learning_rate": 1.923809523809524e-06,
451
+ "logits/chosen": -4.949021339416504,
452
+ "logits/rejected": -5.033829212188721,
453
+ "logps/chosen": -356.3919982910156,
454
+ "logps/rejected": -330.82720947265625,
455
+ "loss": 0.4621857166290283,
456
+ "rewards/accuracies": 0.8999999761581421,
457
+ "rewards/chosen": 0.7969595789909363,
458
+ "rewards/margins": 0.6875919103622437,
459
+ "rewards/rejected": 0.10936765372753143,
460
+ "step": 150
461
+ },
462
+ {
463
+ "epoch": 1.3268817204301075,
464
+ "grad_norm": 1.6490590572357178,
465
+ "learning_rate": 1.8761904761904762e-06,
466
+ "logits/chosen": -5.267385005950928,
467
+ "logits/rejected": -5.265533924102783,
468
+ "logps/chosen": -297.99224853515625,
469
+ "logps/rejected": -326.91339111328125,
470
+ "loss": 0.49877166748046875,
471
+ "rewards/accuracies": 0.800000011920929,
472
+ "rewards/chosen": 0.7068697810173035,
473
+ "rewards/margins": 0.6336223483085632,
474
+ "rewards/rejected": 0.07324743270874023,
475
+ "step": 155
476
+ },
477
+ {
478
+ "epoch": 1.3698924731182796,
479
+ "grad_norm": 1.483849048614502,
480
+ "learning_rate": 1.8285714285714288e-06,
481
+ "logits/chosen": -5.356790065765381,
482
+ "logits/rejected": -5.103802680969238,
483
+ "logps/chosen": -294.2845153808594,
484
+ "logps/rejected": -274.7852783203125,
485
+ "loss": 0.49851350784301757,
486
+ "rewards/accuracies": 0.8500000238418579,
487
+ "rewards/chosen": 0.6102481484413147,
488
+ "rewards/margins": 0.5510420203208923,
489
+ "rewards/rejected": 0.05920610576868057,
490
+ "step": 160
491
+ },
492
+ {
493
+ "epoch": 1.4129032258064516,
494
+ "grad_norm": 1.3165900707244873,
495
+ "learning_rate": 1.780952380952381e-06,
496
+ "logits/chosen": -5.431517124176025,
497
+ "logits/rejected": -5.376145362854004,
498
+ "logps/chosen": -325.1388244628906,
499
+ "logps/rejected": -327.6669616699219,
500
+ "loss": 0.4309373378753662,
501
+ "rewards/accuracies": 0.949999988079071,
502
+ "rewards/chosen": 0.6497762799263,
503
+ "rewards/margins": 0.9170367121696472,
504
+ "rewards/rejected": -0.2672604024410248,
505
+ "step": 165
506
+ },
507
+ {
508
+ "epoch": 1.4559139784946238,
509
+ "grad_norm": 1.5377726554870605,
510
+ "learning_rate": 1.7333333333333332e-06,
511
+ "logits/chosen": -5.27555513381958,
512
+ "logits/rejected": -5.022242069244385,
513
+ "logps/chosen": -340.17779541015625,
514
+ "logps/rejected": -288.0126647949219,
515
+ "loss": 0.4575087547302246,
516
+ "rewards/accuracies": 0.925000011920929,
517
+ "rewards/chosen": 0.5721868276596069,
518
+ "rewards/margins": 0.6390407085418701,
519
+ "rewards/rejected": -0.06685388088226318,
520
+ "step": 170
521
+ },
522
+ {
523
+ "epoch": 1.4989247311827958,
524
+ "grad_norm": 2.0849504470825195,
525
+ "learning_rate": 1.6857142857142858e-06,
526
+ "logits/chosen": -5.070017337799072,
527
+ "logits/rejected": -5.327781677246094,
528
+ "logps/chosen": -360.60809326171875,
529
+ "logps/rejected": -292.8214416503906,
530
+ "loss": 0.45406789779663087,
531
+ "rewards/accuracies": 0.925000011920929,
532
+ "rewards/chosen": 0.748975396156311,
533
+ "rewards/margins": 0.6336467266082764,
534
+ "rewards/rejected": 0.11532865464687347,
535
+ "step": 175
536
+ },
537
+ {
538
+ "epoch": 1.5419354838709678,
539
+ "grad_norm": 1.4946179389953613,
540
+ "learning_rate": 1.638095238095238e-06,
541
+ "logits/chosen": -5.378829002380371,
542
+ "logits/rejected": -5.2273030281066895,
543
+ "logps/chosen": -369.40679931640625,
544
+ "logps/rejected": -310.7534484863281,
545
+ "loss": 0.40073528289794924,
546
+ "rewards/accuracies": 1.0,
547
+ "rewards/chosen": 0.9242110252380371,
548
+ "rewards/margins": 0.8946256637573242,
549
+ "rewards/rejected": 0.029585417360067368,
550
+ "step": 180
551
+ },
552
+ {
553
+ "epoch": 1.5849462365591398,
554
+ "grad_norm": 1.9285597801208496,
555
+ "learning_rate": 1.5904761904761906e-06,
556
+ "logits/chosen": -5.534226417541504,
557
+ "logits/rejected": -5.36181640625,
558
+ "logps/chosen": -285.0791931152344,
559
+ "logps/rejected": -258.01654052734375,
560
+ "loss": 0.44419097900390625,
561
+ "rewards/accuracies": 0.8999999761581421,
562
+ "rewards/chosen": 0.7649332284927368,
563
+ "rewards/margins": 0.6902278661727905,
564
+ "rewards/rejected": 0.07470535486936569,
565
+ "step": 185
566
+ },
567
+ {
568
+ "epoch": 1.6279569892473118,
569
+ "grad_norm": 1.4278947114944458,
570
+ "learning_rate": 1.5428571428571428e-06,
571
+ "logits/chosen": -5.1719889640808105,
572
+ "logits/rejected": -5.186745643615723,
573
+ "logps/chosen": -366.37286376953125,
574
+ "logps/rejected": -310.18048095703125,
575
+ "loss": 0.37736806869506834,
576
+ "rewards/accuracies": 1.0,
577
+ "rewards/chosen": 1.4671168327331543,
578
+ "rewards/margins": 1.110413908958435,
579
+ "rewards/rejected": 0.3567030429840088,
580
+ "step": 190
581
+ },
582
+ {
583
+ "epoch": 1.6709677419354838,
584
+ "grad_norm": 1.7930585145950317,
585
+ "learning_rate": 1.4952380952380954e-06,
586
+ "logits/chosen": -5.458104133605957,
587
+ "logits/rejected": -5.514155864715576,
588
+ "logps/chosen": -288.04083251953125,
589
+ "logps/rejected": -311.49090576171875,
590
+ "loss": 0.43456592559814455,
591
+ "rewards/accuracies": 0.925000011920929,
592
+ "rewards/chosen": 0.7255552411079407,
593
+ "rewards/margins": 0.843337893486023,
594
+ "rewards/rejected": -0.11778266727924347,
595
+ "step": 195
596
+ },
597
+ {
598
+ "epoch": 1.7139784946236558,
599
+ "grad_norm": 1.3665006160736084,
600
+ "learning_rate": 1.4476190476190478e-06,
601
+ "logits/chosen": -5.221610069274902,
602
+ "logits/rejected": -5.051304340362549,
603
+ "logps/chosen": -379.436279296875,
604
+ "logps/rejected": -294.8427734375,
605
+ "loss": 0.3780463218688965,
606
+ "rewards/accuracies": 0.9750000238418579,
607
+ "rewards/chosen": 1.0199247598648071,
608
+ "rewards/margins": 1.0175530910491943,
609
+ "rewards/rejected": 0.0023716867435723543,
610
+ "step": 200
611
+ }
612
+ ],
613
+ "logging_steps": 5,
614
+ "max_steps": 351,
615
+ "num_input_tokens_seen": 0,
616
+ "num_train_epochs": 3,
617
+ "save_steps": 50,
618
+ "stateful_callbacks": {
619
+ "TrainerControl": {
620
+ "args": {
621
+ "should_epoch_stop": false,
622
+ "should_evaluate": false,
623
+ "should_log": false,
624
+ "should_save": true,
625
+ "should_training_stop": false
626
+ },
627
+ "attributes": {}
628
+ }
629
+ },
630
+ "total_flos": 0.0,
631
+ "train_batch_size": 1,
632
+ "trial_name": null,
633
+ "trial_params": null
634
+ }
gemma-9b-dpo/checkpoint-200/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74ee585106626f1196ffc9833586c5c11146fc384e5b33ed522cf45bab148032
3
+ size 6097
gemma-9b-dpo/checkpoint-250/README.md ADDED
@@ -0,0 +1,209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: google/gemma-2-9b-it
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - base_model:adapter:google/gemma-2-9b-it
7
+ - dpo
8
+ - lora
9
+ - transformers
10
+ - trl
11
+ ---
12
+
13
+ # Model Card for Model ID
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+
19
+ ## Model Details
20
+
21
+ ### Model Description
22
+
23
+ <!-- Provide a longer summary of what this model is. -->
24
+
25
+
26
+
27
+ - **Developed by:** [More Information Needed]
28
+ - **Funded by [optional]:** [More Information Needed]
29
+ - **Shared by [optional]:** [More Information Needed]
30
+ - **Model type:** [More Information Needed]
31
+ - **Language(s) (NLP):** [More Information Needed]
32
+ - **License:** [More Information Needed]
33
+ - **Finetuned from model [optional]:** [More Information Needed]
34
+
35
+ ### Model Sources [optional]
36
+
37
+ <!-- Provide the basic links for the model. -->
38
+
39
+ - **Repository:** [More Information Needed]
40
+ - **Paper [optional]:** [More Information Needed]
41
+ - **Demo [optional]:** [More Information Needed]
42
+
43
+ ## Uses
44
+
45
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
46
+
47
+ ### Direct Use
48
+
49
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
50
+
51
+ [More Information Needed]
52
+
53
+ ### Downstream Use [optional]
54
+
55
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
56
+
57
+ [More Information Needed]
58
+
59
+ ### Out-of-Scope Use
60
+
61
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
62
+
63
+ [More Information Needed]
64
+
65
+ ## Bias, Risks, and Limitations
66
+
67
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
68
+
69
+ [More Information Needed]
70
+
71
+ ### Recommendations
72
+
73
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
74
+
75
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
76
+
77
+ ## How to Get Started with the Model
78
+
79
+ Use the code below to get started with the model.
80
+
81
+ [More Information Needed]
82
+
83
+ ## Training Details
84
+
85
+ ### Training Data
86
+
87
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
88
+
89
+ [More Information Needed]
90
+
91
+ ### Training Procedure
92
+
93
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
94
+
95
+ #### Preprocessing [optional]
96
+
97
+ [More Information Needed]
98
+
99
+
100
+ #### Training Hyperparameters
101
+
102
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
103
+
104
+ #### Speeds, Sizes, Times [optional]
105
+
106
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
107
+
108
+ [More Information Needed]
109
+
110
+ ## Evaluation
111
+
112
+ <!-- This section describes the evaluation protocols and provides the results. -->
113
+
114
+ ### Testing Data, Factors & Metrics
115
+
116
+ #### Testing Data
117
+
118
+ <!-- This should link to a Dataset Card if possible. -->
119
+
120
+ [More Information Needed]
121
+
122
+ #### Factors
123
+
124
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
125
+
126
+ [More Information Needed]
127
+
128
+ #### Metrics
129
+
130
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
131
+
132
+ [More Information Needed]
133
+
134
+ ### Results
135
+
136
+ [More Information Needed]
137
+
138
+ #### Summary
139
+
140
+
141
+
142
+ ## Model Examination [optional]
143
+
144
+ <!-- Relevant interpretability work for the model goes here -->
145
+
146
+ [More Information Needed]
147
+
148
+ ## Environmental Impact
149
+
150
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
151
+
152
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
153
+
154
+ - **Hardware Type:** [More Information Needed]
155
+ - **Hours used:** [More Information Needed]
156
+ - **Cloud Provider:** [More Information Needed]
157
+ - **Compute Region:** [More Information Needed]
158
+ - **Carbon Emitted:** [More Information Needed]
159
+
160
+ ## Technical Specifications [optional]
161
+
162
+ ### Model Architecture and Objective
163
+
164
+ [More Information Needed]
165
+
166
+ ### Compute Infrastructure
167
+
168
+ [More Information Needed]
169
+
170
+ #### Hardware
171
+
172
+ [More Information Needed]
173
+
174
+ #### Software
175
+
176
+ [More Information Needed]
177
+
178
+ ## Citation [optional]
179
+
180
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
181
+
182
+ **BibTeX:**
183
+
184
+ [More Information Needed]
185
+
186
+ **APA:**
187
+
188
+ [More Information Needed]
189
+
190
+ ## Glossary [optional]
191
+
192
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
193
+
194
+ [More Information Needed]
195
+
196
+ ## More Information [optional]
197
+
198
+ [More Information Needed]
199
+
200
+ ## Model Card Authors [optional]
201
+
202
+ [More Information Needed]
203
+
204
+ ## Model Card Contact
205
+
206
+ [More Information Needed]
207
+ ### Framework versions
208
+
209
+ - PEFT 0.18.1
gemma-9b-dpo/checkpoint-250/adapter_config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "google/gemma-2-9b-it",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": false,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 32,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.05,
22
+ "megatron_config": null,
23
+ "megatron_core": "megatron.core",
24
+ "modules_to_save": null,
25
+ "peft_type": "LORA",
26
+ "peft_version": "0.18.1",
27
+ "qalora_group_size": 16,
28
+ "r": 16,
29
+ "rank_pattern": {},
30
+ "revision": null,
31
+ "target_modules": [
32
+ "down_proj",
33
+ "v_proj",
34
+ "k_proj",
35
+ "up_proj",
36
+ "q_proj",
37
+ "gate_proj",
38
+ "o_proj"
39
+ ],
40
+ "target_parameters": null,
41
+ "task_type": "CAUSAL_LM",
42
+ "trainable_token_indices": null,
43
+ "use_dora": false,
44
+ "use_qalora": false,
45
+ "use_rslora": false
46
+ }
gemma-9b-dpo/checkpoint-250/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:704ff1ccb6a1f6beee1d50c669cc8f81d0fbcbb03b1dd612f4e299ec66230ff9
3
+ size 216151256
gemma-9b-dpo/checkpoint-250/chat_template.jinja ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {{ bos_token }}{% if messages[0]['role'] == 'system' %}{{ raise_exception('System role not supported') }}{% endif %}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if (message['role'] == 'assistant') %}{% set role = 'model' %}{% else %}{% set role = message['role'] %}{% endif %}{{ '<start_of_turn>' + role + '
2
+ ' + message['content'] | trim + '<end_of_turn>
3
+ ' }}{% endfor %}{% if add_generation_prompt %}{{'<start_of_turn>model
4
+ '}}{% endif %}
gemma-9b-dpo/checkpoint-250/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3573a3bb365b29c41a0cf112e05c12b594cdec400b68371a2726c2d28574913e
3
+ size 110425877
gemma-9b-dpo/checkpoint-250/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2534e434cd5abbb8f7668d3eab0549db0ef95d6a797a3efa86b712e8e32266a7
3
+ size 14645
gemma-9b-dpo/checkpoint-250/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f12b49b9980a9d56fac92a1aed3dfea4f671a19e8cfb1c75b50d9e52b6848d73
3
+ size 1465
gemma-9b-dpo/checkpoint-250/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:394ace002a144ac6ad5486387502f2d36f70c087310c3d907857240c76fcb36e
3
+ size 34362748
gemma-9b-dpo/checkpoint-250/tokenizer_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "<bos>",
4
+ "clean_up_tokenization_spaces": false,
5
+ "eos_token": "<eos>",
6
+ "extra_special_tokens": [
7
+ "<start_of_turn>",
8
+ "<end_of_turn>"
9
+ ],
10
+ "is_local": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 1000000000000000019884624838656,
13
+ "pad_token": "<pad>",
14
+ "sp_model_kwargs": {},
15
+ "spaces_between_special_tokens": false,
16
+ "tokenizer_class": "GemmaTokenizer",
17
+ "unk_token": "<unk>",
18
+ "use_default_system_prompt": false
19
+ }
gemma-9b-dpo/checkpoint-250/trainer_state.json ADDED
@@ -0,0 +1,784 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.1376344086021506,
6
+ "eval_steps": 500,
7
+ "global_step": 250,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.043010752688172046,
14
+ "grad_norm": 1.6562875509262085,
15
+ "learning_rate": 3.333333333333333e-07,
16
+ "logits/chosen": -5.94815731048584,
17
+ "logits/rejected": -5.856410503387451,
18
+ "logps/chosen": -302.07794189453125,
19
+ "logps/rejected": -283.5802001953125,
20
+ "loss": 0.6852486610412598,
21
+ "rewards/accuracies": 0.375,
22
+ "rewards/chosen": 0.01349079143255949,
23
+ "rewards/margins": 0.017757166177034378,
24
+ "rewards/rejected": -0.00426637614145875,
25
+ "step": 5
26
+ },
27
+ {
28
+ "epoch": 0.08602150537634409,
29
+ "grad_norm": 1.7723957300186157,
30
+ "learning_rate": 7.5e-07,
31
+ "logits/chosen": -5.585428714752197,
32
+ "logits/rejected": -5.577895641326904,
33
+ "logps/chosen": -362.89239501953125,
34
+ "logps/rejected": -280.8010559082031,
35
+ "loss": 0.6844330310821534,
36
+ "rewards/accuracies": 0.6499999761581421,
37
+ "rewards/chosen": 0.018745478242635727,
38
+ "rewards/margins": 0.018739622086286545,
39
+ "rewards/rejected": 5.85438692723983e-06,
40
+ "step": 10
41
+ },
42
+ {
43
+ "epoch": 0.12903225806451613,
44
+ "grad_norm": 1.6630758047103882,
45
+ "learning_rate": 1.1666666666666668e-06,
46
+ "logits/chosen": -5.916059494018555,
47
+ "logits/rejected": -5.797668933868408,
48
+ "logps/chosen": -307.4602966308594,
49
+ "logps/rejected": -316.79803466796875,
50
+ "loss": 0.7070020198822021,
51
+ "rewards/accuracies": 0.5249999761581421,
52
+ "rewards/chosen": -0.008706416003406048,
53
+ "rewards/margins": -0.02138950116932392,
54
+ "rewards/rejected": 0.012683087959885597,
55
+ "step": 15
56
+ },
57
+ {
58
+ "epoch": 0.17204301075268819,
59
+ "grad_norm": 5.225494861602783,
60
+ "learning_rate": 1.5833333333333333e-06,
61
+ "logits/chosen": -5.818185329437256,
62
+ "logits/rejected": -5.638014793395996,
63
+ "logps/chosen": -329.3303527832031,
64
+ "logps/rejected": -292.2645568847656,
65
+ "loss": 0.6929943561553955,
66
+ "rewards/accuracies": 0.4749999940395355,
67
+ "rewards/chosen": -0.044999174773693085,
68
+ "rewards/margins": 0.00216915225610137,
69
+ "rewards/rejected": -0.04716832935810089,
70
+ "step": 20
71
+ },
72
+ {
73
+ "epoch": 0.21505376344086022,
74
+ "grad_norm": 2.5229244232177734,
75
+ "learning_rate": 2e-06,
76
+ "logits/chosen": -5.8189592361450195,
77
+ "logits/rejected": -5.646462917327881,
78
+ "logps/chosen": -311.9922180175781,
79
+ "logps/rejected": -297.0098571777344,
80
+ "loss": 0.7002316951751709,
81
+ "rewards/accuracies": 0.4000000059604645,
82
+ "rewards/chosen": 0.0544702522456646,
83
+ "rewards/margins": -0.012132339179515839,
84
+ "rewards/rejected": 0.06660258769989014,
85
+ "step": 25
86
+ },
87
+ {
88
+ "epoch": 0.25806451612903225,
89
+ "grad_norm": 1.6524324417114258,
90
+ "learning_rate": 2.4166666666666667e-06,
91
+ "logits/chosen": -5.696784496307373,
92
+ "logits/rejected": -5.6943678855896,
93
+ "logps/chosen": -350.2938537597656,
94
+ "logps/rejected": -322.43182373046875,
95
+ "loss": 0.699475908279419,
96
+ "rewards/accuracies": 0.574999988079071,
97
+ "rewards/chosen": -0.0020792910363525152,
98
+ "rewards/margins": 0.02348965033888817,
99
+ "rewards/rejected": -0.025568943470716476,
100
+ "step": 30
101
+ },
102
+ {
103
+ "epoch": 0.3010752688172043,
104
+ "grad_norm": 9.075493812561035,
105
+ "learning_rate": 2.8333333333333335e-06,
106
+ "logits/chosen": -5.407717704772949,
107
+ "logits/rejected": -5.5642805099487305,
108
+ "logps/chosen": -391.3837585449219,
109
+ "logps/rejected": -333.9232482910156,
110
+ "loss": 0.6926415920257568,
111
+ "rewards/accuracies": 0.4749999940395355,
112
+ "rewards/chosen": -0.026233959943056107,
113
+ "rewards/margins": 0.00582819152623415,
114
+ "rewards/rejected": -0.03206215053796768,
115
+ "step": 35
116
+ },
117
+ {
118
+ "epoch": 0.34408602150537637,
119
+ "grad_norm": 2.6282100677490234,
120
+ "learning_rate": 2.9714285714285716e-06,
121
+ "logits/chosen": -5.649778366088867,
122
+ "logits/rejected": -5.512002944946289,
123
+ "logps/chosen": -338.31048583984375,
124
+ "logps/rejected": -279.06536865234375,
125
+ "loss": 0.6739111423492432,
126
+ "rewards/accuracies": 0.625,
127
+ "rewards/chosen": 0.04578382521867752,
128
+ "rewards/margins": 0.042618148028850555,
129
+ "rewards/rejected": 0.0031656839419156313,
130
+ "step": 40
131
+ },
132
+ {
133
+ "epoch": 0.3870967741935484,
134
+ "grad_norm": 16.96689796447754,
135
+ "learning_rate": 2.923809523809524e-06,
136
+ "logits/chosen": -5.518254280090332,
137
+ "logits/rejected": -5.577255725860596,
138
+ "logps/chosen": -374.29449462890625,
139
+ "logps/rejected": -366.4861145019531,
140
+ "loss": 0.7162120819091797,
141
+ "rewards/accuracies": 0.550000011920929,
142
+ "rewards/chosen": 0.03416462987661362,
143
+ "rewards/margins": 0.0036566159687936306,
144
+ "rewards/rejected": 0.030508000403642654,
145
+ "step": 45
146
+ },
147
+ {
148
+ "epoch": 0.43010752688172044,
149
+ "grad_norm": 1.447816252708435,
150
+ "learning_rate": 2.8761904761904764e-06,
151
+ "logits/chosen": -5.831571102142334,
152
+ "logits/rejected": -5.802765846252441,
153
+ "logps/chosen": -316.0965881347656,
154
+ "logps/rejected": -269.0494384765625,
155
+ "loss": 0.6838803291320801,
156
+ "rewards/accuracies": 0.550000011920929,
157
+ "rewards/chosen": -0.1265859305858612,
158
+ "rewards/margins": 0.026753634214401245,
159
+ "rewards/rejected": -0.15333956480026245,
160
+ "step": 50
161
+ },
162
+ {
163
+ "epoch": 0.4731182795698925,
164
+ "grad_norm": 2.174459457397461,
165
+ "learning_rate": 2.8285714285714288e-06,
166
+ "logits/chosen": -5.777364253997803,
167
+ "logits/rejected": -5.834509372711182,
168
+ "logps/chosen": -316.31390380859375,
169
+ "logps/rejected": -282.5376892089844,
170
+ "loss": 0.6832016944885254,
171
+ "rewards/accuracies": 0.675000011920929,
172
+ "rewards/chosen": -0.01756533980369568,
173
+ "rewards/margins": 0.02713041380047798,
174
+ "rewards/rejected": -0.04469575732946396,
175
+ "step": 55
176
+ },
177
+ {
178
+ "epoch": 0.5161290322580645,
179
+ "grad_norm": 2.2876009941101074,
180
+ "learning_rate": 2.780952380952381e-06,
181
+ "logits/chosen": -5.92340087890625,
182
+ "logits/rejected": -5.846226692199707,
183
+ "logps/chosen": -235.49667358398438,
184
+ "logps/rejected": -267.7764892578125,
185
+ "loss": 0.6656109809875488,
186
+ "rewards/accuracies": 0.6499999761581421,
187
+ "rewards/chosen": 0.021378064528107643,
188
+ "rewards/margins": 0.06001085042953491,
189
+ "rewards/rejected": -0.038632798939943314,
190
+ "step": 60
191
+ },
192
+ {
193
+ "epoch": 0.5591397849462365,
194
+ "grad_norm": 3.45564603805542,
195
+ "learning_rate": 2.733333333333333e-06,
196
+ "logits/chosen": -5.650258541107178,
197
+ "logits/rejected": -5.6721577644348145,
198
+ "logps/chosen": -381.40655517578125,
199
+ "logps/rejected": -311.68865966796875,
200
+ "loss": 0.7090614795684814,
201
+ "rewards/accuracies": 0.4749999940395355,
202
+ "rewards/chosen": 0.1167660504579544,
203
+ "rewards/margins": -0.0002812861348502338,
204
+ "rewards/rejected": 0.11704733222723007,
205
+ "step": 65
206
+ },
207
+ {
208
+ "epoch": 0.6021505376344086,
209
+ "grad_norm": 4.502689838409424,
210
+ "learning_rate": 2.685714285714286e-06,
211
+ "logits/chosen": -5.727735996246338,
212
+ "logits/rejected": -5.644078254699707,
213
+ "logps/chosen": -362.89886474609375,
214
+ "logps/rejected": -314.7126770019531,
215
+ "loss": 0.6596640110015869,
216
+ "rewards/accuracies": 0.6499999761581421,
217
+ "rewards/chosen": 0.07285688072443008,
218
+ "rewards/margins": 0.07270809262990952,
219
+ "rewards/rejected": 0.00014879256195854396,
220
+ "step": 70
221
+ },
222
+ {
223
+ "epoch": 0.6451612903225806,
224
+ "grad_norm": 2.2681005001068115,
225
+ "learning_rate": 2.638095238095238e-06,
226
+ "logits/chosen": -5.528594017028809,
227
+ "logits/rejected": -5.29849910736084,
228
+ "logps/chosen": -353.0726013183594,
229
+ "logps/rejected": -331.8568115234375,
230
+ "loss": 0.7387234687805175,
231
+ "rewards/accuracies": 0.574999988079071,
232
+ "rewards/chosen": 0.1685911864042282,
233
+ "rewards/margins": -0.04706493765115738,
234
+ "rewards/rejected": 0.2156561315059662,
235
+ "step": 75
236
+ },
237
+ {
238
+ "epoch": 0.6881720430107527,
239
+ "grad_norm": 3.0135886669158936,
240
+ "learning_rate": 2.5904761904761907e-06,
241
+ "logits/chosen": -5.8677144050598145,
242
+ "logits/rejected": -5.5384016036987305,
243
+ "logps/chosen": -326.17724609375,
244
+ "logps/rejected": -285.8974304199219,
245
+ "loss": 0.6047093391418457,
246
+ "rewards/accuracies": 0.925000011920929,
247
+ "rewards/chosen": 0.19864055514335632,
248
+ "rewards/margins": 0.19678126275539398,
249
+ "rewards/rejected": 0.0018593042623251677,
250
+ "step": 80
251
+ },
252
+ {
253
+ "epoch": 0.7311827956989247,
254
+ "grad_norm": 1.7684988975524902,
255
+ "learning_rate": 2.5428571428571427e-06,
256
+ "logits/chosen": -5.413943290710449,
257
+ "logits/rejected": -5.427316665649414,
258
+ "logps/chosen": -372.1170654296875,
259
+ "logps/rejected": -312.30279541015625,
260
+ "loss": 0.651799201965332,
261
+ "rewards/accuracies": 0.675000011920929,
262
+ "rewards/chosen": 0.179380863904953,
263
+ "rewards/margins": 0.09846550226211548,
264
+ "rewards/rejected": 0.08091535419225693,
265
+ "step": 85
266
+ },
267
+ {
268
+ "epoch": 0.7741935483870968,
269
+ "grad_norm": 2.4072341918945312,
270
+ "learning_rate": 2.4952380952380955e-06,
271
+ "logits/chosen": -5.4685492515563965,
272
+ "logits/rejected": -5.51275634765625,
273
+ "logps/chosen": -345.37188720703125,
274
+ "logps/rejected": -273.87261962890625,
275
+ "loss": 0.6362700462341309,
276
+ "rewards/accuracies": 0.6499999761581421,
277
+ "rewards/chosen": 0.33122482895851135,
278
+ "rewards/margins": 0.1338074505329132,
279
+ "rewards/rejected": 0.19741736352443695,
280
+ "step": 90
281
+ },
282
+ {
283
+ "epoch": 0.8172043010752689,
284
+ "grad_norm": 1.745698094367981,
285
+ "learning_rate": 2.4476190476190475e-06,
286
+ "logits/chosen": -5.6481757164001465,
287
+ "logits/rejected": -5.4260029792785645,
288
+ "logps/chosen": -337.11767578125,
289
+ "logps/rejected": -263.3458251953125,
290
+ "loss": 0.6222721099853515,
291
+ "rewards/accuracies": 0.824999988079071,
292
+ "rewards/chosen": 0.25297099351882935,
293
+ "rewards/margins": 0.1701255589723587,
294
+ "rewards/rejected": 0.08284540474414825,
295
+ "step": 95
296
+ },
297
+ {
298
+ "epoch": 0.8602150537634409,
299
+ "grad_norm": 1.4592756032943726,
300
+ "learning_rate": 2.4000000000000003e-06,
301
+ "logits/chosen": -5.402789115905762,
302
+ "logits/rejected": -5.280846118927002,
303
+ "logps/chosen": -304.55316162109375,
304
+ "logps/rejected": -314.95306396484375,
305
+ "loss": 0.6404934883117676,
306
+ "rewards/accuracies": 0.699999988079071,
307
+ "rewards/chosen": 0.29103565216064453,
308
+ "rewards/margins": 0.12328717857599258,
309
+ "rewards/rejected": 0.16774848103523254,
310
+ "step": 100
311
+ },
312
+ {
313
+ "epoch": 0.9032258064516129,
314
+ "grad_norm": 7.651180744171143,
315
+ "learning_rate": 2.3523809523809523e-06,
316
+ "logits/chosen": -5.333284854888916,
317
+ "logits/rejected": -5.209356307983398,
318
+ "logps/chosen": -385.76483154296875,
319
+ "logps/rejected": -317.0154724121094,
320
+ "loss": 0.5562876224517822,
321
+ "rewards/accuracies": 0.824999988079071,
322
+ "rewards/chosen": 0.5905097126960754,
323
+ "rewards/margins": 0.39645156264305115,
324
+ "rewards/rejected": 0.19405809044837952,
325
+ "step": 105
326
+ },
327
+ {
328
+ "epoch": 0.946236559139785,
329
+ "grad_norm": 2.2162137031555176,
330
+ "learning_rate": 2.304761904761905e-06,
331
+ "logits/chosen": -5.9164533615112305,
332
+ "logits/rejected": -5.660351276397705,
333
+ "logps/chosen": -362.57720947265625,
334
+ "logps/rejected": -328.386962890625,
335
+ "loss": 0.5672832489013672,
336
+ "rewards/accuracies": 0.824999988079071,
337
+ "rewards/chosen": 0.4752024710178375,
338
+ "rewards/margins": 0.3182791471481323,
339
+ "rewards/rejected": 0.15692326426506042,
340
+ "step": 110
341
+ },
342
+ {
343
+ "epoch": 0.989247311827957,
344
+ "grad_norm": 2.102505683898926,
345
+ "learning_rate": 2.257142857142857e-06,
346
+ "logits/chosen": -5.367539882659912,
347
+ "logits/rejected": -5.304169654846191,
348
+ "logps/chosen": -389.9139404296875,
349
+ "logps/rejected": -303.767822265625,
350
+ "loss": 0.563088321685791,
351
+ "rewards/accuracies": 0.800000011920929,
352
+ "rewards/chosen": 0.9107279777526855,
353
+ "rewards/margins": 0.7586480379104614,
354
+ "rewards/rejected": 0.15207989513874054,
355
+ "step": 115
356
+ },
357
+ {
358
+ "epoch": 1.0258064516129033,
359
+ "grad_norm": 2.9593698978424072,
360
+ "learning_rate": 2.20952380952381e-06,
361
+ "logits/chosen": -5.506618976593018,
362
+ "logits/rejected": -5.349832057952881,
363
+ "logps/chosen": -299.4593505859375,
364
+ "logps/rejected": -229.88784790039062,
365
+ "loss": 0.5665022850036621,
366
+ "rewards/accuracies": 0.8823529481887817,
367
+ "rewards/chosen": 0.4371771514415741,
368
+ "rewards/margins": 0.3876599073410034,
369
+ "rewards/rejected": 0.04951724037528038,
370
+ "step": 120
371
+ },
372
+ {
373
+ "epoch": 1.0688172043010753,
374
+ "grad_norm": 2.5988082885742188,
375
+ "learning_rate": 2.161904761904762e-06,
376
+ "logits/chosen": -5.542575836181641,
377
+ "logits/rejected": -5.428658485412598,
378
+ "logps/chosen": -353.8345642089844,
379
+ "logps/rejected": -312.9586486816406,
380
+ "loss": 0.5137276172637939,
381
+ "rewards/accuracies": 0.925000011920929,
382
+ "rewards/chosen": 0.6325557827949524,
383
+ "rewards/margins": 0.44065460562705994,
384
+ "rewards/rejected": 0.19190113246440887,
385
+ "step": 125
386
+ },
387
+ {
388
+ "epoch": 1.1118279569892473,
389
+ "grad_norm": 1.6948280334472656,
390
+ "learning_rate": 2.1142857142857147e-06,
391
+ "logits/chosen": -5.847611427307129,
392
+ "logits/rejected": -5.498036861419678,
393
+ "logps/chosen": -300.35577392578125,
394
+ "logps/rejected": -311.05126953125,
395
+ "loss": 0.5773736953735351,
396
+ "rewards/accuracies": 0.875,
397
+ "rewards/chosen": 0.6476074457168579,
398
+ "rewards/margins": 0.37814000248908997,
399
+ "rewards/rejected": 0.2694675028324127,
400
+ "step": 130
401
+ },
402
+ {
403
+ "epoch": 1.1548387096774193,
404
+ "grad_norm": 1.372768759727478,
405
+ "learning_rate": 2.0666666666666666e-06,
406
+ "logits/chosen": -5.619741916656494,
407
+ "logits/rejected": -5.618372917175293,
408
+ "logps/chosen": -367.06732177734375,
409
+ "logps/rejected": -301.2478332519531,
410
+ "loss": 0.46464052200317385,
411
+ "rewards/accuracies": 0.949999988079071,
412
+ "rewards/chosen": 0.8353813290596008,
413
+ "rewards/margins": 0.6073431968688965,
414
+ "rewards/rejected": 0.22803807258605957,
415
+ "step": 135
416
+ },
417
+ {
418
+ "epoch": 1.1978494623655913,
419
+ "grad_norm": 2.498854637145996,
420
+ "learning_rate": 2.0190476190476195e-06,
421
+ "logits/chosen": -5.195496559143066,
422
+ "logits/rejected": -4.919422149658203,
423
+ "logps/chosen": -333.09979248046875,
424
+ "logps/rejected": -317.78167724609375,
425
+ "loss": 0.5423533916473389,
426
+ "rewards/accuracies": 0.8500000238418579,
427
+ "rewards/chosen": 0.46827277541160583,
428
+ "rewards/margins": 0.42974653840065,
429
+ "rewards/rejected": 0.0385262668132782,
430
+ "step": 140
431
+ },
432
+ {
433
+ "epoch": 1.2408602150537635,
434
+ "grad_norm": 1.451978087425232,
435
+ "learning_rate": 1.9714285714285714e-06,
436
+ "logits/chosen": -5.425684452056885,
437
+ "logits/rejected": -5.322096824645996,
438
+ "logps/chosen": -314.4623107910156,
439
+ "logps/rejected": -271.85479736328125,
440
+ "loss": 0.4632129192352295,
441
+ "rewards/accuracies": 0.949999988079071,
442
+ "rewards/chosen": 0.9266977310180664,
443
+ "rewards/margins": 0.672536313533783,
444
+ "rewards/rejected": 0.25416144728660583,
445
+ "step": 145
446
+ },
447
+ {
448
+ "epoch": 1.2838709677419355,
449
+ "grad_norm": 2.0581016540527344,
450
+ "learning_rate": 1.923809523809524e-06,
451
+ "logits/chosen": -4.949021339416504,
452
+ "logits/rejected": -5.033829212188721,
453
+ "logps/chosen": -356.3919982910156,
454
+ "logps/rejected": -330.82720947265625,
455
+ "loss": 0.4621857166290283,
456
+ "rewards/accuracies": 0.8999999761581421,
457
+ "rewards/chosen": 0.7969595789909363,
458
+ "rewards/margins": 0.6875919103622437,
459
+ "rewards/rejected": 0.10936765372753143,
460
+ "step": 150
461
+ },
462
+ {
463
+ "epoch": 1.3268817204301075,
464
+ "grad_norm": 1.6490590572357178,
465
+ "learning_rate": 1.8761904761904762e-06,
466
+ "logits/chosen": -5.267385005950928,
467
+ "logits/rejected": -5.265533924102783,
468
+ "logps/chosen": -297.99224853515625,
469
+ "logps/rejected": -326.91339111328125,
470
+ "loss": 0.49877166748046875,
471
+ "rewards/accuracies": 0.800000011920929,
472
+ "rewards/chosen": 0.7068697810173035,
473
+ "rewards/margins": 0.6336223483085632,
474
+ "rewards/rejected": 0.07324743270874023,
475
+ "step": 155
476
+ },
477
+ {
478
+ "epoch": 1.3698924731182796,
479
+ "grad_norm": 1.483849048614502,
480
+ "learning_rate": 1.8285714285714288e-06,
481
+ "logits/chosen": -5.356790065765381,
482
+ "logits/rejected": -5.103802680969238,
483
+ "logps/chosen": -294.2845153808594,
484
+ "logps/rejected": -274.7852783203125,
485
+ "loss": 0.49851350784301757,
486
+ "rewards/accuracies": 0.8500000238418579,
487
+ "rewards/chosen": 0.6102481484413147,
488
+ "rewards/margins": 0.5510420203208923,
489
+ "rewards/rejected": 0.05920610576868057,
490
+ "step": 160
491
+ },
492
+ {
493
+ "epoch": 1.4129032258064516,
494
+ "grad_norm": 1.3165900707244873,
495
+ "learning_rate": 1.780952380952381e-06,
496
+ "logits/chosen": -5.431517124176025,
497
+ "logits/rejected": -5.376145362854004,
498
+ "logps/chosen": -325.1388244628906,
499
+ "logps/rejected": -327.6669616699219,
500
+ "loss": 0.4309373378753662,
501
+ "rewards/accuracies": 0.949999988079071,
502
+ "rewards/chosen": 0.6497762799263,
503
+ "rewards/margins": 0.9170367121696472,
504
+ "rewards/rejected": -0.2672604024410248,
505
+ "step": 165
506
+ },
507
+ {
508
+ "epoch": 1.4559139784946238,
509
+ "grad_norm": 1.5377726554870605,
510
+ "learning_rate": 1.7333333333333332e-06,
511
+ "logits/chosen": -5.27555513381958,
512
+ "logits/rejected": -5.022242069244385,
513
+ "logps/chosen": -340.17779541015625,
514
+ "logps/rejected": -288.0126647949219,
515
+ "loss": 0.4575087547302246,
516
+ "rewards/accuracies": 0.925000011920929,
517
+ "rewards/chosen": 0.5721868276596069,
518
+ "rewards/margins": 0.6390407085418701,
519
+ "rewards/rejected": -0.06685388088226318,
520
+ "step": 170
521
+ },
522
+ {
523
+ "epoch": 1.4989247311827958,
524
+ "grad_norm": 2.0849504470825195,
525
+ "learning_rate": 1.6857142857142858e-06,
526
+ "logits/chosen": -5.070017337799072,
527
+ "logits/rejected": -5.327781677246094,
528
+ "logps/chosen": -360.60809326171875,
529
+ "logps/rejected": -292.8214416503906,
530
+ "loss": 0.45406789779663087,
531
+ "rewards/accuracies": 0.925000011920929,
532
+ "rewards/chosen": 0.748975396156311,
533
+ "rewards/margins": 0.6336467266082764,
534
+ "rewards/rejected": 0.11532865464687347,
535
+ "step": 175
536
+ },
537
+ {
538
+ "epoch": 1.5419354838709678,
539
+ "grad_norm": 1.4946179389953613,
540
+ "learning_rate": 1.638095238095238e-06,
541
+ "logits/chosen": -5.378829002380371,
542
+ "logits/rejected": -5.2273030281066895,
543
+ "logps/chosen": -369.40679931640625,
544
+ "logps/rejected": -310.7534484863281,
545
+ "loss": 0.40073528289794924,
546
+ "rewards/accuracies": 1.0,
547
+ "rewards/chosen": 0.9242110252380371,
548
+ "rewards/margins": 0.8946256637573242,
549
+ "rewards/rejected": 0.029585417360067368,
550
+ "step": 180
551
+ },
552
+ {
553
+ "epoch": 1.5849462365591398,
554
+ "grad_norm": 1.9285597801208496,
555
+ "learning_rate": 1.5904761904761906e-06,
556
+ "logits/chosen": -5.534226417541504,
557
+ "logits/rejected": -5.36181640625,
558
+ "logps/chosen": -285.0791931152344,
559
+ "logps/rejected": -258.01654052734375,
560
+ "loss": 0.44419097900390625,
561
+ "rewards/accuracies": 0.8999999761581421,
562
+ "rewards/chosen": 0.7649332284927368,
563
+ "rewards/margins": 0.6902278661727905,
564
+ "rewards/rejected": 0.07470535486936569,
565
+ "step": 185
566
+ },
567
+ {
568
+ "epoch": 1.6279569892473118,
569
+ "grad_norm": 1.4278947114944458,
570
+ "learning_rate": 1.5428571428571428e-06,
571
+ "logits/chosen": -5.1719889640808105,
572
+ "logits/rejected": -5.186745643615723,
573
+ "logps/chosen": -366.37286376953125,
574
+ "logps/rejected": -310.18048095703125,
575
+ "loss": 0.37736806869506834,
576
+ "rewards/accuracies": 1.0,
577
+ "rewards/chosen": 1.4671168327331543,
578
+ "rewards/margins": 1.110413908958435,
579
+ "rewards/rejected": 0.3567030429840088,
580
+ "step": 190
581
+ },
582
+ {
583
+ "epoch": 1.6709677419354838,
584
+ "grad_norm": 1.7930585145950317,
585
+ "learning_rate": 1.4952380952380954e-06,
586
+ "logits/chosen": -5.458104133605957,
587
+ "logits/rejected": -5.514155864715576,
588
+ "logps/chosen": -288.04083251953125,
589
+ "logps/rejected": -311.49090576171875,
590
+ "loss": 0.43456592559814455,
591
+ "rewards/accuracies": 0.925000011920929,
592
+ "rewards/chosen": 0.7255552411079407,
593
+ "rewards/margins": 0.843337893486023,
594
+ "rewards/rejected": -0.11778266727924347,
595
+ "step": 195
596
+ },
597
+ {
598
+ "epoch": 1.7139784946236558,
599
+ "grad_norm": 1.3665006160736084,
600
+ "learning_rate": 1.4476190476190478e-06,
601
+ "logits/chosen": -5.221610069274902,
602
+ "logits/rejected": -5.051304340362549,
603
+ "logps/chosen": -379.436279296875,
604
+ "logps/rejected": -294.8427734375,
605
+ "loss": 0.3780463218688965,
606
+ "rewards/accuracies": 0.9750000238418579,
607
+ "rewards/chosen": 1.0199247598648071,
608
+ "rewards/margins": 1.0175530910491943,
609
+ "rewards/rejected": 0.0023716867435723543,
610
+ "step": 200
611
+ },
612
+ {
613
+ "epoch": 1.7569892473118278,
614
+ "grad_norm": 1.2653045654296875,
615
+ "learning_rate": 1.4000000000000001e-06,
616
+ "logits/chosen": -5.128909111022949,
617
+ "logits/rejected": -4.9968366622924805,
618
+ "logps/chosen": -374.5747375488281,
619
+ "logps/rejected": -310.3750915527344,
620
+ "loss": 0.42014646530151367,
621
+ "rewards/accuracies": 0.925000011920929,
622
+ "rewards/chosen": 1.18593430519104,
623
+ "rewards/margins": 1.1055948734283447,
624
+ "rewards/rejected": 0.08033928275108337,
625
+ "step": 205
626
+ },
627
+ {
628
+ "epoch": 1.8,
629
+ "grad_norm": 1.0835349559783936,
630
+ "learning_rate": 1.3523809523809525e-06,
631
+ "logits/chosen": -5.398374557495117,
632
+ "logits/rejected": -5.446703910827637,
633
+ "logps/chosen": -332.9021911621094,
634
+ "logps/rejected": -307.1310119628906,
635
+ "loss": 0.4112071990966797,
636
+ "rewards/accuracies": 0.8999999761581421,
637
+ "rewards/chosen": 1.0410544872283936,
638
+ "rewards/margins": 1.0919562578201294,
639
+ "rewards/rejected": -0.05090172961354256,
640
+ "step": 210
641
+ },
642
+ {
643
+ "epoch": 1.843010752688172,
644
+ "grad_norm": 1.8526785373687744,
645
+ "learning_rate": 1.3047619047619047e-06,
646
+ "logits/chosen": -5.290182113647461,
647
+ "logits/rejected": -5.226934909820557,
648
+ "logps/chosen": -376.863525390625,
649
+ "logps/rejected": -320.2511291503906,
650
+ "loss": 0.4023551940917969,
651
+ "rewards/accuracies": 0.949999988079071,
652
+ "rewards/chosen": 1.0466766357421875,
653
+ "rewards/margins": 0.915771484375,
654
+ "rewards/rejected": 0.1309051215648651,
655
+ "step": 215
656
+ },
657
+ {
658
+ "epoch": 1.886021505376344,
659
+ "grad_norm": 1.6254379749298096,
660
+ "learning_rate": 1.2571428571428571e-06,
661
+ "logits/chosen": -5.433383464813232,
662
+ "logits/rejected": -5.424604892730713,
663
+ "logps/chosen": -276.48382568359375,
664
+ "logps/rejected": -254.852294921875,
665
+ "loss": 0.42627677917480467,
666
+ "rewards/accuracies": 0.925000011920929,
667
+ "rewards/chosen": 1.1158783435821533,
668
+ "rewards/margins": 0.8330133557319641,
669
+ "rewards/rejected": 0.2828650176525116,
670
+ "step": 220
671
+ },
672
+ {
673
+ "epoch": 1.9290322580645163,
674
+ "grad_norm": 3.121758460998535,
675
+ "learning_rate": 1.2095238095238095e-06,
676
+ "logits/chosen": -5.485299110412598,
677
+ "logits/rejected": -5.462108612060547,
678
+ "logps/chosen": -276.477294921875,
679
+ "logps/rejected": -285.85516357421875,
680
+ "loss": 0.4259671688079834,
681
+ "rewards/accuracies": 0.8999999761581421,
682
+ "rewards/chosen": 0.8818387985229492,
683
+ "rewards/margins": 0.7649229764938354,
684
+ "rewards/rejected": 0.11691585928201675,
685
+ "step": 225
686
+ },
687
+ {
688
+ "epoch": 1.9720430107526883,
689
+ "grad_norm": 2.3801393508911133,
690
+ "learning_rate": 1.161904761904762e-06,
691
+ "logits/chosen": -5.211794376373291,
692
+ "logits/rejected": -4.747314929962158,
693
+ "logps/chosen": -416.1683654785156,
694
+ "logps/rejected": -365.8246765136719,
695
+ "loss": 0.3910404443740845,
696
+ "rewards/accuracies": 0.925000011920929,
697
+ "rewards/chosen": 1.5805695056915283,
698
+ "rewards/margins": 1.2381842136383057,
699
+ "rewards/rejected": 0.34238511323928833,
700
+ "step": 230
701
+ },
702
+ {
703
+ "epoch": 2.0086021505376346,
704
+ "grad_norm": 1.0709431171417236,
705
+ "learning_rate": 1.1142857142857143e-06,
706
+ "logits/chosen": -5.255922794342041,
707
+ "logits/rejected": -5.191451549530029,
708
+ "logps/chosen": -329.1349182128906,
709
+ "logps/rejected": -281.3226623535156,
710
+ "loss": 0.4125385761260986,
711
+ "rewards/accuracies": 0.970588207244873,
712
+ "rewards/chosen": 1.2619037628173828,
713
+ "rewards/margins": 1.028172254562378,
714
+ "rewards/rejected": 0.2337314933538437,
715
+ "step": 235
716
+ },
717
+ {
718
+ "epoch": 2.0516129032258066,
719
+ "grad_norm": 1.131330966949463,
720
+ "learning_rate": 1.0666666666666667e-06,
721
+ "logits/chosen": -5.735346794128418,
722
+ "logits/rejected": -5.384338855743408,
723
+ "logps/chosen": -354.9601135253906,
724
+ "logps/rejected": -285.20147705078125,
725
+ "loss": 0.3143571138381958,
726
+ "rewards/accuracies": 1.0,
727
+ "rewards/chosen": 1.6434673070907593,
728
+ "rewards/margins": 1.7338101863861084,
729
+ "rewards/rejected": -0.09034281969070435,
730
+ "step": 240
731
+ },
732
+ {
733
+ "epoch": 2.0946236559139786,
734
+ "grad_norm": 1.3222469091415405,
735
+ "learning_rate": 1.019047619047619e-06,
736
+ "logits/chosen": -5.4371256828308105,
737
+ "logits/rejected": -5.41799259185791,
738
+ "logps/chosen": -318.0994567871094,
739
+ "logps/rejected": -296.23126220703125,
740
+ "loss": 0.3331931114196777,
741
+ "rewards/accuracies": 1.0,
742
+ "rewards/chosen": 0.9413619041442871,
743
+ "rewards/margins": 1.1318610906600952,
744
+ "rewards/rejected": -0.1904991865158081,
745
+ "step": 245
746
+ },
747
+ {
748
+ "epoch": 2.1376344086021506,
749
+ "grad_norm": 1.1309553384780884,
750
+ "learning_rate": 9.714285714285715e-07,
751
+ "logits/chosen": -5.32918643951416,
752
+ "logits/rejected": -5.228451251983643,
753
+ "logps/chosen": -386.4266662597656,
754
+ "logps/rejected": -299.82586669921875,
755
+ "loss": 0.30677978992462157,
756
+ "rewards/accuracies": 0.9750000238418579,
757
+ "rewards/chosen": 1.5374553203582764,
758
+ "rewards/margins": 1.2936350107192993,
759
+ "rewards/rejected": 0.24382023513317108,
760
+ "step": 250
761
+ }
762
+ ],
763
+ "logging_steps": 5,
764
+ "max_steps": 351,
765
+ "num_input_tokens_seen": 0,
766
+ "num_train_epochs": 3,
767
+ "save_steps": 50,
768
+ "stateful_callbacks": {
769
+ "TrainerControl": {
770
+ "args": {
771
+ "should_epoch_stop": false,
772
+ "should_evaluate": false,
773
+ "should_log": false,
774
+ "should_save": true,
775
+ "should_training_stop": false
776
+ },
777
+ "attributes": {}
778
+ }
779
+ },
780
+ "total_flos": 0.0,
781
+ "train_batch_size": 1,
782
+ "trial_name": null,
783
+ "trial_params": null
784
+ }
gemma-9b-dpo/checkpoint-250/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74ee585106626f1196ffc9833586c5c11146fc384e5b33ed522cf45bab148032
3
+ size 6097
gemma-9b-dpo/checkpoint-300/README.md ADDED
@@ -0,0 +1,209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: google/gemma-2-9b-it
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - base_model:adapter:google/gemma-2-9b-it
7
+ - dpo
8
+ - lora
9
+ - transformers
10
+ - trl
11
+ ---
12
+
13
+ # Model Card for Model ID
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+
19
+ ## Model Details
20
+
21
+ ### Model Description
22
+
23
+ <!-- Provide a longer summary of what this model is. -->
24
+
25
+
26
+
27
+ - **Developed by:** [More Information Needed]
28
+ - **Funded by [optional]:** [More Information Needed]
29
+ - **Shared by [optional]:** [More Information Needed]
30
+ - **Model type:** [More Information Needed]
31
+ - **Language(s) (NLP):** [More Information Needed]
32
+ - **License:** [More Information Needed]
33
+ - **Finetuned from model [optional]:** [More Information Needed]
34
+
35
+ ### Model Sources [optional]
36
+
37
+ <!-- Provide the basic links for the model. -->
38
+
39
+ - **Repository:** [More Information Needed]
40
+ - **Paper [optional]:** [More Information Needed]
41
+ - **Demo [optional]:** [More Information Needed]
42
+
43
+ ## Uses
44
+
45
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
46
+
47
+ ### Direct Use
48
+
49
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
50
+
51
+ [More Information Needed]
52
+
53
+ ### Downstream Use [optional]
54
+
55
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
56
+
57
+ [More Information Needed]
58
+
59
+ ### Out-of-Scope Use
60
+
61
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
62
+
63
+ [More Information Needed]
64
+
65
+ ## Bias, Risks, and Limitations
66
+
67
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
68
+
69
+ [More Information Needed]
70
+
71
+ ### Recommendations
72
+
73
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
74
+
75
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
76
+
77
+ ## How to Get Started with the Model
78
+
79
+ Use the code below to get started with the model.
80
+
81
+ [More Information Needed]
82
+
83
+ ## Training Details
84
+
85
+ ### Training Data
86
+
87
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
88
+
89
+ [More Information Needed]
90
+
91
+ ### Training Procedure
92
+
93
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
94
+
95
+ #### Preprocessing [optional]
96
+
97
+ [More Information Needed]
98
+
99
+
100
+ #### Training Hyperparameters
101
+
102
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
103
+
104
+ #### Speeds, Sizes, Times [optional]
105
+
106
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
107
+
108
+ [More Information Needed]
109
+
110
+ ## Evaluation
111
+
112
+ <!-- This section describes the evaluation protocols and provides the results. -->
113
+
114
+ ### Testing Data, Factors & Metrics
115
+
116
+ #### Testing Data
117
+
118
+ <!-- This should link to a Dataset Card if possible. -->
119
+
120
+ [More Information Needed]
121
+
122
+ #### Factors
123
+
124
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
125
+
126
+ [More Information Needed]
127
+
128
+ #### Metrics
129
+
130
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
131
+
132
+ [More Information Needed]
133
+
134
+ ### Results
135
+
136
+ [More Information Needed]
137
+
138
+ #### Summary
139
+
140
+
141
+
142
+ ## Model Examination [optional]
143
+
144
+ <!-- Relevant interpretability work for the model goes here -->
145
+
146
+ [More Information Needed]
147
+
148
+ ## Environmental Impact
149
+
150
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
151
+
152
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
153
+
154
+ - **Hardware Type:** [More Information Needed]
155
+ - **Hours used:** [More Information Needed]
156
+ - **Cloud Provider:** [More Information Needed]
157
+ - **Compute Region:** [More Information Needed]
158
+ - **Carbon Emitted:** [More Information Needed]
159
+
160
+ ## Technical Specifications [optional]
161
+
162
+ ### Model Architecture and Objective
163
+
164
+ [More Information Needed]
165
+
166
+ ### Compute Infrastructure
167
+
168
+ [More Information Needed]
169
+
170
+ #### Hardware
171
+
172
+ [More Information Needed]
173
+
174
+ #### Software
175
+
176
+ [More Information Needed]
177
+
178
+ ## Citation [optional]
179
+
180
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
181
+
182
+ **BibTeX:**
183
+
184
+ [More Information Needed]
185
+
186
+ **APA:**
187
+
188
+ [More Information Needed]
189
+
190
+ ## Glossary [optional]
191
+
192
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
193
+
194
+ [More Information Needed]
195
+
196
+ ## More Information [optional]
197
+
198
+ [More Information Needed]
199
+
200
+ ## Model Card Authors [optional]
201
+
202
+ [More Information Needed]
203
+
204
+ ## Model Card Contact
205
+
206
+ [More Information Needed]
207
+ ### Framework versions
208
+
209
+ - PEFT 0.18.1
gemma-9b-dpo/checkpoint-300/adapter_config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "google/gemma-2-9b-it",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": false,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 32,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.05,
22
+ "megatron_config": null,
23
+ "megatron_core": "megatron.core",
24
+ "modules_to_save": null,
25
+ "peft_type": "LORA",
26
+ "peft_version": "0.18.1",
27
+ "qalora_group_size": 16,
28
+ "r": 16,
29
+ "rank_pattern": {},
30
+ "revision": null,
31
+ "target_modules": [
32
+ "down_proj",
33
+ "v_proj",
34
+ "k_proj",
35
+ "up_proj",
36
+ "q_proj",
37
+ "gate_proj",
38
+ "o_proj"
39
+ ],
40
+ "target_parameters": null,
41
+ "task_type": "CAUSAL_LM",
42
+ "trainable_token_indices": null,
43
+ "use_dora": false,
44
+ "use_qalora": false,
45
+ "use_rslora": false
46
+ }