Hugging Face's logo

superiort
/
T3Q-LLM-sft1.0-dpo1.0_4300QA_10epochs

PEFT
Safetensors
llama
axolotl
Generated from Trainer
Model card Files Community
superiort commited on
Commit
8382899
•
1 Parent(s): 0c6078b

End of training

Browse files
Files changed (2) hide show
  1. README.md +263 -0
  2. adapter_model.bin +3 -0
README.md ADDED
@@ -0,0 +1,263 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: peft
4
+ tags:
5
+ - axolotl
6
+ - generated_from_trainer
7
+ base_model: T3Q-LLM/T3Q-LLM-sft1.0-dpo1.0
8
+ model-index:
9
+ - name: T3Q-LLM-sft1.0-dpo1.0_4300QA
10
+ results: []
11
+ ---
12
+
13
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
14
+ should probably proofread and complete it, then remove this comment. -->
15
+
16
+ [<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
17
+ <details><summary>See axolotl config</summary>
18
+
19
+ axolotl version: `0.4.0`
20
+ ```yaml
21
+ base_model: T3Q-LLM/T3Q-LLM-sft1.0-dpo1.0
22
+ base_model_config: T3Q-LLM/T3Q-LLM-sft1.0-dpo1.0
23
+ model_type: AutoModelForCausalLM
24
+ tokenizer_type: AutoTokenizer
25
+ is_llama_derived_model: true
26
+ hub_model_id: T3Q-LLM-sft1.0-dpo1.0_4300QA
27
+
28
+ load_in_8bit: false
29
+ load_in_4bit: true
30
+ strict: false
31
+
32
+ datasets:
33
+ # - path: admin_data.csv
34
+ - path: superiort/multiplechoice-4300
35
+ type: alpaca
36
+ # The below are defaults. only set what's needed if you use a different column name.
37
+ # system_prompt: ""
38
+ # system_format: "{system}"
39
+ # field_system: system
40
+ # field_instruction: instruction
41
+ # field_input: input
42
+ # field_output: output
43
+
44
+ # format: |-
45
+ # Human: {instruction} {input}
46
+ # Assistant:
47
+
48
+ # no_input_format: "{instruction} "
49
+
50
+ # dataset_prepared_path: yanolja_preprocessed_data
51
+ dataset_prepared_path: last_run_prepared
52
+ val_set_size: 0.2
53
+ output_dir: ./T3Q-LLM-sft1.0-dpo1.0_4300QA
54
+
55
+ adapter: qlora
56
+ lora_model_dir:
57
+
58
+ # device_map: [0,1,3]
59
+
60
+ sequence_len: 4096
61
+ sample_packing: false
62
+
63
+ lora_r: 32
64
+ lora_alpha: 16
65
+ lora_dropout: 0.05
66
+ lora_target_modules:
67
+ lora_target_linear: true
68
+ lora_fan_in_fan_out:
69
+
70
+ wandb_project: axolotl_T3Q_4300
71
+ wandb_entity:
72
+ wandb_watch:
73
+ wandb_run_id: T3Q_mod_4300
74
+ wandb_log_model:
75
+
76
+ gradient_accumulation_steps: 4
77
+ micro_batch_size: 2
78
+ num_epochs: 10
79
+ optimizer: paged_adamw_32bit
80
+ lr_scheduler: cosine
81
+ learning_rate: 0.0002
82
+
83
+ train_on_inputs: false
84
+ group_by_length: false
85
+ bf16: true
86
+ fp16: false
87
+ tf32: false
88
+
89
+ gradient_checkpointing: true
90
+ early_stopping_patience:
91
+ resume_from_checkpoint:
92
+ local_rank:
93
+ logging_steps: 1
94
+ xformers_attention:
95
+ flash_attention: true
96
+
97
+ warmup_steps: 100
98
+ eval_steps: 0.01
99
+ save_strategy: epoch
100
+ save_steps:
101
+ debug:
102
+ deepspeed:
103
+ weight_decay: 0.0
104
+ fsdp:
105
+ fsdp_config:
106
+ special_tokens:
107
+ bos_token: "<s>"
108
+ eos_token: "<|im_end|>"
109
+ unk_token: "<unk>"
110
+ pad_token: "</s>" # EOS와 PAD가 동일
111
+
112
+ ```
113
+
114
+ </details><br>
115
+
116
+ # T3Q-LLM-sft1.0-dpo1.0_4300QA
117
+
118
+ This model is a fine-tuned version of [T3Q-LLM/T3Q-LLM-sft1.0-dpo1.0](https://huggingface.co/T3Q-LLM/T3Q-LLM-sft1.0-dpo1.0) on the None dataset.
119
+ It achieves the following results on the evaluation set:
120
+ - Loss: 1.2288
121
+
122
+ ## Model description
123
+
124
+ More information needed
125
+
126
+ ## Intended uses & limitations
127
+
128
+ More information needed
129
+
130
+ ## Training and evaluation data
131
+
132
+ More information needed
133
+
134
+ ## Training procedure
135
+
136
+ ### Training hyperparameters
137
+
138
+ The following hyperparameters were used during training:
139
+ - learning_rate: 0.0002
140
+ - train_batch_size: 2
141
+ - eval_batch_size: 2
142
+ - seed: 42
143
+ - distributed_type: multi-GPU
144
+ - num_devices: 4
145
+ - gradient_accumulation_steps: 4
146
+ - total_train_batch_size: 32
147
+ - total_eval_batch_size: 8
148
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
149
+ - lr_scheduler_type: cosine
150
+ - lr_scheduler_warmup_steps: 100
151
+ - num_epochs: 10
152
+
153
+ ### Training results
154
+
155
+ | Training Loss | Epoch | Step | Validation Loss |
156
+ |:-------------:|:------:|:----:|:---------------:|
157
+ | 1.2424 | 0.0093 | 1 | 1.0432 |
158
+ | 1.0333 | 0.1023 | 11 | 0.9004 |
159
+ | 0.8715 | 0.2047 | 22 | 0.7157 |
160
+ | 0.7053 | 0.3070 | 33 | 0.6548 |
161
+ | 0.6688 | 0.4093 | 44 | 0.6449 |
162
+ | 0.6823 | 0.5116 | 55 | 0.6282 |
163
+ | 0.5876 | 0.6140 | 66 | 0.6251 |
164
+ | 0.6994 | 0.7163 | 77 | 0.6290 |
165
+ | 0.6662 | 0.8186 | 88 | 0.6311 |
166
+ | 0.6239 | 0.9209 | 99 | 0.6338 |
167
+ | 0.5959 | 1.0233 | 110 | 0.6319 |
168
+ | 0.6408 | 1.1256 | 121 | 0.6668 |
169
+ | 0.595 | 1.2279 | 132 | 0.6221 |
170
+ | 0.5476 | 1.3302 | 143 | 0.6295 |
171
+ | 0.587 | 1.4326 | 154 | 0.6569 |
172
+ | 0.5867 | 1.5349 | 165 | 0.6208 |
173
+ | 0.5895 | 1.6372 | 176 | 0.6264 |
174
+ | 0.6581 | 1.7395 | 187 | 0.6208 |
175
+ | 0.5872 | 1.8419 | 198 | 0.6290 |
176
+ | 0.6314 | 1.9442 | 209 | 0.6243 |
177
+ | 0.4397 | 2.0465 | 220 | 0.6591 |
178
+ | 0.4568 | 2.1488 | 231 | 0.7095 |
179
+ | 0.422 | 2.2512 | 242 | 0.6914 |
180
+ | 0.453 | 2.3535 | 253 | 0.7001 |
181
+ | 0.4678 | 2.4558 | 264 | 0.6896 |
182
+ | 0.4335 | 2.5581 | 275 | 0.6776 |
183
+ | 0.4796 | 2.6605 | 286 | 0.6829 |
184
+ | 0.4637 | 2.7628 | 297 | 0.6742 |
185
+ | 0.4532 | 2.8651 | 308 | 0.6828 |
186
+ | 0.4348 | 2.9674 | 319 | 0.6836 |
187
+ | 0.2787 | 3.0698 | 330 | 0.8085 |
188
+ | 0.2336 | 3.1721 | 341 | 0.8380 |
189
+ | 0.2341 | 3.2744 | 352 | 0.7998 |
190
+ | 0.2393 | 3.3767 | 363 | 0.8041 |
191
+ | 0.2826 | 3.4791 | 374 | 0.8040 |
192
+ | 0.2505 | 3.5814 | 385 | 0.8099 |
193
+ | 0.3057 | 3.6837 | 396 | 0.8103 |
194
+ | 0.2789 | 3.7860 | 407 | 0.7964 |
195
+ | 0.269 | 3.8884 | 418 | 0.7891 |
196
+ | 0.2493 | 3.9907 | 429 | 0.7958 |
197
+ | 0.1193 | 4.0930 | 440 | 0.9242 |
198
+ | 0.1143 | 4.1953 | 451 | 0.9331 |
199
+ | 0.1147 | 4.2977 | 462 | 0.9112 |
200
+ | 0.1351 | 4.4 | 473 | 0.9290 |
201
+ | 0.0982 | 4.5023 | 484 | 0.9358 |
202
+ | 0.1011 | 4.6047 | 495 | 0.9279 |
203
+ | 0.09 | 4.7070 | 506 | 0.9289 |
204
+ | 0.1063 | 4.8093 | 517 | 0.9392 |
205
+ | 0.1038 | 4.9116 | 528 | 0.9267 |
206
+ | 0.0361 | 5.0140 | 539 | 0.9412 |
207
+ | 0.0371 | 5.1163 | 550 | 1.0589 |
208
+ | 0.033 | 5.2186 | 561 | 1.0253 |
209
+ | 0.0426 | 5.3209 | 572 | 1.0482 |
210
+ | 0.0357 | 5.4233 | 583 | 1.0388 |
211
+ | 0.0355 | 5.5256 | 594 | 1.0566 |
212
+ | 0.0373 | 5.6279 | 605 | 1.0470 |
213
+ | 0.0395 | 5.7302 | 616 | 1.0581 |
214
+ | 0.0366 | 5.8326 | 627 | 1.0696 |
215
+ | 0.0387 | 5.9349 | 638 | 1.0641 |
216
+ | 0.0127 | 6.0372 | 649 | 1.0692 |
217
+ | 0.0114 | 6.1395 | 660 | 1.1612 |
218
+ | 0.0105 | 6.2419 | 671 | 1.1575 |
219
+ | 0.0121 | 6.3442 | 682 | 1.1479 |
220
+ | 0.0082 | 6.4465 | 693 | 1.1591 |
221
+ | 0.011 | 6.5488 | 704 | 1.1669 |
222
+ | 0.0112 | 6.6512 | 715 | 1.1645 |
223
+ | 0.0109 | 6.7535 | 726 | 1.1628 |
224
+ | 0.0102 | 6.8558 | 737 | 1.1705 |
225
+ | 0.0098 | 6.9581 | 748 | 1.1769 |
226
+ | 0.006 | 7.0605 | 759 | 1.1840 |
227
+ | 0.0064 | 7.1628 | 770 | 1.2016 |
228
+ | 0.0063 | 7.2651 | 781 | 1.2133 |
229
+ | 0.0058 | 7.3674 | 792 | 1.2182 |
230
+ | 0.0056 | 7.4698 | 803 | 1.2218 |
231
+ | 0.0057 | 7.5721 | 814 | 1.2234 |
232
+ | 0.0059 | 7.6744 | 825 | 1.2245 |
233
+ | 0.0057 | 7.7767 | 836 | 1.2247 |
234
+ | 0.0048 | 7.8791 | 847 | 1.2247 |
235
+ | 0.0054 | 7.9814 | 858 | 1.2246 |
236
+ | 0.0051 | 8.0837 | 869 | 1.2252 |
237
+ | 0.0059 | 8.1860 | 880 | 1.2261 |
238
+ | 0.0053 | 8.2884 | 891 | 1.2272 |
239
+ | 0.0057 | 8.3907 | 902 | 1.2275 |
240
+ | 0.0056 | 8.4930 | 913 | 1.2280 |
241
+ | 0.0052 | 8.5953 | 924 | 1.2283 |
242
+ | 0.007 | 8.6977 | 935 | 1.2287 |
243
+ | 0.0052 | 8.8 | 946 | 1.2285 |
244
+ | 0.005 | 8.9023 | 957 | 1.2289 |
245
+ | 0.0056 | 9.0047 | 968 | 1.2288 |
246
+ | 0.005 | 9.1070 | 979 | 1.2289 |
247
+ | 0.0054 | 9.2093 | 990 | 1.2290 |
248
+ | 0.0053 | 9.3116 | 1001 | 1.2288 |
249
+ | 0.0049 | 9.4140 | 1012 | 1.2290 |
250
+ | 0.0052 | 9.5163 | 1023 | 1.2290 |
251
+ | 0.0058 | 9.6186 | 1034 | 1.2291 |
252
+ | 0.0059 | 9.7209 | 1045 | 1.2289 |
253
+ | 0.0055 | 9.8233 | 1056 | 1.2289 |
254
+ | 0.0054 | 9.9256 | 1067 | 1.2288 |
255
+
256
+
257
+ ### Framework versions
258
+
259
+ - PEFT 0.10.0
260
+ - Transformers 4.40.1
261
+ - Pytorch 2.1.2+cu121
262
+ - Datasets 2.15.0
263
+ - Tokenizers 0.19.1
adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c8b5dd44eebbdd932fda643d6b86c880bd1005e64260ac48fd496399b571610f
3
+ size 251901130