Hugging Face's logo

chchen
/
Mistral-7B-Instruct-v0.3-ORPO

PEFT
Safetensors
llama-factory
lora
trl
dpo
Generated from Trainer
Model card Files
xet
Community
chchen commited on
Commit
e8ca2bf
·
verified ·
1 Parent(s): 373951e

Training in progress, step 500

Browse files
Files changed (10) hide show
  1. .ipynb_checkpoints/lora_orpo-checkpoint.yaml +43 -0
  2. adapter_config.json +34 -0
  3. adapter_model.safetensors +3 -0
  4. lora_orpo.yaml +43 -0
  5. special_tokens_map.json +24 -0
  6. tokenizer.json +0 -0
  7. tokenizer.model +3 -0
  8. tokenizer_config.json +0 -0
  9. trainer_log.jsonl +51 -0
  10. training_args.bin +3 -0
.ipynb_checkpoints/lora_orpo-checkpoint.yaml ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ### model
2
+ model_name_or_path: mistralai/Mistral-7B-Instruct-v0.3
3
+
4
+ ### method
5
+ stage: orpo
6
+ do_train: true
7
+ finetuning_type: lora
8
+ lora_target: all
9
+
10
+ ### dataset
11
+ dataset: dpo_mix_en
12
+ dataset_dir: data
13
+ template: mistral
14
+ cutoff_len: 1024
15
+ # max_samples: 1000
16
+ overwrite_cache: true
17
+ preprocessing_num_workers: 16
18
+
19
+ ### output
20
+ output_dir: saves/Mistral-7B-Instruct-v0.3/lora/orpo
21
+ logging_steps: 10
22
+ save_steps: 500
23
+ plot_loss: true
24
+ overwrite_output_dir: true
25
+ save_total_limit: 3
26
+ load_best_model_at_end: true
27
+ push_to_hub: true
28
+ hub_model_id: chchen/Mistral-7B-Instruct-v0.3-ORPO
29
+
30
+ ### train
31
+ per_device_train_batch_size: 2
32
+ gradient_accumulation_steps: 8
33
+ learning_rate: 0.000005
34
+ num_train_epochs: 3.0
35
+ lr_scheduler_type: cosine
36
+ warmup_steps: 0.1
37
+ bf16: true
38
+
39
+ ### eval
40
+ val_size: 0.1
41
+ per_device_eval_batch_size: 2
42
+ evaluation_strategy: steps
43
+ eval_steps: 500
adapter_config.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "mistralai/Mistral-7B-Instruct-v0.3",
5
+ "bias": "none",
6
+ "fan_in_fan_out": false,
7
+ "inference_mode": true,
8
+ "init_lora_weights": true,
9
+ "layer_replication": null,
10
+ "layers_pattern": null,
11
+ "layers_to_transform": null,
12
+ "loftq_config": {},
13
+ "lora_alpha": 16,
14
+ "lora_dropout": 0.0,
15
+ "megatron_config": null,
16
+ "megatron_core": "megatron.core",
17
+ "modules_to_save": null,
18
+ "peft_type": "LORA",
19
+ "r": 8,
20
+ "rank_pattern": {},
21
+ "revision": null,
22
+ "target_modules": [
23
+ "gate_proj",
24
+ "up_proj",
25
+ "o_proj",
26
+ "v_proj",
27
+ "q_proj",
28
+ "down_proj",
29
+ "k_proj"
30
+ ],
31
+ "task_type": "CAUSAL_LM",
32
+ "use_dora": false,
33
+ "use_rslora": false
34
+ }
adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09a8acfbf35a8a366284f3658253aebf69472fe237cef71ffcef54976eb0d738
3
+ size 83945296
lora_orpo.yaml ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ### model
2
+ model_name_or_path: mistralai/Mistral-7B-Instruct-v0.3
3
+
4
+ ### method
5
+ stage: orpo
6
+ do_train: true
7
+ finetuning_type: lora
8
+ lora_target: all
9
+
10
+ ### dataset
11
+ dataset: dpo_mix_en
12
+ dataset_dir: data
13
+ template: mistral
14
+ cutoff_len: 1024
15
+ # max_samples: 1000
16
+ overwrite_cache: true
17
+ preprocessing_num_workers: 16
18
+
19
+ ### output
20
+ output_dir: saves/Mistral-7B-Instruct-v0.3/lora/orpo
21
+ logging_steps: 10
22
+ save_steps: 500
23
+ plot_loss: true
24
+ overwrite_output_dir: true
25
+ save_total_limit: 3
26
+ load_best_model_at_end: true
27
+ push_to_hub: true
28
+ hub_model_id: chchen/Mistral-7B-Instruct-v0.3-ORPO
29
+
30
+ ### train
31
+ per_device_train_batch_size: 2
32
+ gradient_accumulation_steps: 8
33
+ learning_rate: 0.000005
34
+ num_train_epochs: 3.0
35
+ lr_scheduler_type: cosine
36
+ warmup_steps: 0.1
37
+ bf16: true
38
+
39
+ ### eval
40
+ val_size: 0.1
41
+ per_device_eval_batch_size: 2
42
+ evaluation_strategy: steps
43
+ eval_steps: 500
special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "</s>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37f00374dea48658ee8f5d0f21895b9bc55cb0103939607c8185bfd1c6ca1f89
3
+ size 587404
tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff
 
trainer_log.jsonl ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"current_steps": 10, "total_steps": 1686, "loss": 1.0773, "accuracy": 0.518750011920929, "learning_rate": 4.9995745934141085e-06, "epoch": 0.017781729273171815, "percentage": 0.59, "elapsed_time": "0:02:57", "remaining_time": "8:15:19"}
2
+ {"current_steps": 20, "total_steps": 1686, "loss": 1.1014, "accuracy": 0.5687500238418579, "learning_rate": 4.9982812903243405e-06, "epoch": 0.03556345854634363, "percentage": 1.19, "elapsed_time": "0:05:56", "remaining_time": "8:15:27"}
3
+ {"current_steps": 30, "total_steps": 1686, "loss": 1.0616, "accuracy": 0.6000000238418579, "learning_rate": 4.996120496405222e-06, "epoch": 0.05334518781951545, "percentage": 1.78, "elapsed_time": "0:08:53", "remaining_time": "8:11:16"}
4
+ {"current_steps": 40, "total_steps": 1686, "loss": 1.1331, "accuracy": 0.5375000238418579, "learning_rate": 4.99309296196014e-06, "epoch": 0.07112691709268726, "percentage": 2.37, "elapsed_time": "0:11:49", "remaining_time": "8:06:42"}
5
+ {"current_steps": 50, "total_steps": 1686, "loss": 0.9797, "accuracy": 0.6499999761581421, "learning_rate": 4.989199738255166e-06, "epoch": 0.08890864636585907, "percentage": 2.97, "elapsed_time": "0:14:49", "remaining_time": "8:05:17"}
6
+ {"current_steps": 60, "total_steps": 1686, "loss": 1.0632, "accuracy": 0.550000011920929, "learning_rate": 4.984442177154031e-06, "epoch": 0.1066903756390309, "percentage": 3.56, "elapsed_time": "0:17:54", "remaining_time": "8:05:05"}
7
+ {"current_steps": 70, "total_steps": 1686, "loss": 1.0027, "accuracy": 0.5562499761581421, "learning_rate": 4.978821930648704e-06, "epoch": 0.12447210491220272, "percentage": 4.15, "elapsed_time": "0:21:00", "remaining_time": "8:04:48"}
8
+ {"current_steps": 80, "total_steps": 1686, "loss": 0.9964, "accuracy": 0.550000011920929, "learning_rate": 4.97234095028576e-06, "epoch": 0.14225383418537452, "percentage": 4.74, "elapsed_time": "0:24:06", "remaining_time": "8:03:55"}
9
+ {"current_steps": 90, "total_steps": 1686, "loss": 0.9472, "accuracy": 0.606249988079071, "learning_rate": 4.965001486488743e-06, "epoch": 0.16003556345854633, "percentage": 5.34, "elapsed_time": "0:27:08", "remaining_time": "8:01:18"}
10
+ {"current_steps": 100, "total_steps": 1686, "loss": 0.9857, "accuracy": 0.5874999761581421, "learning_rate": 4.956806087776732e-06, "epoch": 0.17781729273171815, "percentage": 5.93, "elapsed_time": "0:30:02", "remaining_time": "7:56:34"}
11
+ {"current_steps": 110, "total_steps": 1686, "loss": 1.0259, "accuracy": 0.581250011920929, "learning_rate": 4.947757599879411e-06, "epoch": 0.19559902200489, "percentage": 6.52, "elapsed_time": "0:32:54", "remaining_time": "7:51:29"}
12
+ {"current_steps": 120, "total_steps": 1686, "loss": 0.9473, "accuracy": 0.5687500238418579, "learning_rate": 4.937859164748931e-06, "epoch": 0.2133807512780618, "percentage": 7.12, "elapsed_time": "0:35:46", "remaining_time": "7:46:52"}
13
+ {"current_steps": 130, "total_steps": 1686, "loss": 0.9558, "accuracy": 0.5562499761581421, "learning_rate": 4.92711421946891e-06, "epoch": 0.23116248055123362, "percentage": 7.71, "elapsed_time": "0:38:56", "remaining_time": "7:46:08"}
14
+ {"current_steps": 140, "total_steps": 1686, "loss": 0.9238, "accuracy": 0.5687500238418579, "learning_rate": 4.915526495060961e-06, "epoch": 0.24894420982440543, "percentage": 8.3, "elapsed_time": "0:42:00", "remaining_time": "7:43:51"}
15
+ {"current_steps": 150, "total_steps": 1686, "loss": 0.9062, "accuracy": 0.606249988079071, "learning_rate": 4.903100015189153e-06, "epoch": 0.26672593909757725, "percentage": 8.9, "elapsed_time": "0:45:03", "remaining_time": "7:41:23"}
16
+ {"current_steps": 160, "total_steps": 1686, "loss": 0.952, "accuracy": 0.5687500238418579, "learning_rate": 4.889839094762848e-06, "epoch": 0.28450766837074903, "percentage": 9.49, "elapsed_time": "0:47:59", "remaining_time": "7:37:38"}
17
+ {"current_steps": 170, "total_steps": 1686, "loss": 0.9602, "accuracy": 0.543749988079071, "learning_rate": 4.875748338438416e-06, "epoch": 0.3022893976439209, "percentage": 10.08, "elapsed_time": "0:50:53", "remaining_time": "7:33:52"}
18
+ {"current_steps": 180, "total_steps": 1686, "loss": 0.9188, "accuracy": 0.5687500238418579, "learning_rate": 4.8608326390203386e-06, "epoch": 0.32007112691709266, "percentage": 10.68, "elapsed_time": "0:53:54", "remaining_time": "7:31:00"}
19
+ {"current_steps": 190, "total_steps": 1686, "loss": 0.9743, "accuracy": 0.543749988079071, "learning_rate": 4.845097175762251e-06, "epoch": 0.3378528561902645, "percentage": 11.27, "elapsed_time": "0:56:54", "remaining_time": "7:28:06"}
20
+ {"current_steps": 200, "total_steps": 1686, "loss": 0.9757, "accuracy": 0.53125, "learning_rate": 4.8285474125685286e-06, "epoch": 0.3556345854634363, "percentage": 11.86, "elapsed_time": "0:59:51", "remaining_time": "7:24:41"}
21
+ {"current_steps": 210, "total_steps": 1686, "loss": 0.9473, "accuracy": 0.5562499761581421, "learning_rate": 4.811189096097025e-06, "epoch": 0.37341631473660813, "percentage": 12.46, "elapsed_time": "1:03:03", "remaining_time": "7:23:14"}
22
+ {"current_steps": 220, "total_steps": 1686, "loss": 0.9507, "accuracy": 0.581250011920929, "learning_rate": 4.793028253763633e-06, "epoch": 0.39119804400978, "percentage": 13.05, "elapsed_time": "1:06:04", "remaining_time": "7:20:21"}
23
+ {"current_steps": 230, "total_steps": 1686, "loss": 0.9107, "accuracy": 0.581250011920929, "learning_rate": 4.774071191649352e-06, "epoch": 0.40897977328295176, "percentage": 13.64, "elapsed_time": "1:09:09", "remaining_time": "7:17:50"}
24
+ {"current_steps": 240, "total_steps": 1686, "loss": 0.9829, "accuracy": 0.518750011920929, "learning_rate": 4.7543244923105975e-06, "epoch": 0.4267615025561236, "percentage": 14.23, "elapsed_time": "1:12:12", "remaining_time": "7:15:01"}
25
+ {"current_steps": 250, "total_steps": 1686, "loss": 0.9739, "accuracy": 0.550000011920929, "learning_rate": 4.733795012493506e-06, "epoch": 0.4445432318292954, "percentage": 14.83, "elapsed_time": "1:15:16", "remaining_time": "7:12:20"}
26
+ {"current_steps": 260, "total_steps": 1686, "loss": 0.8776, "accuracy": 0.5625, "learning_rate": 4.712489880753035e-06, "epoch": 0.46232496110246724, "percentage": 15.42, "elapsed_time": "1:18:24", "remaining_time": "7:10:02"}
27
+ {"current_steps": 270, "total_steps": 1686, "loss": 0.8925, "accuracy": 0.6000000238418579, "learning_rate": 4.690416494977673e-06, "epoch": 0.480106690375639, "percentage": 16.01, "elapsed_time": "1:21:25", "remaining_time": "7:07:03"}
28
+ {"current_steps": 280, "total_steps": 1686, "loss": 1.018, "accuracy": 0.5687500238418579, "learning_rate": 4.667582519820639e-06, "epoch": 0.49788841964881086, "percentage": 16.61, "elapsed_time": "1:24:37", "remaining_time": "7:04:55"}
29
+ {"current_steps": 290, "total_steps": 1686, "loss": 0.9404, "accuracy": 0.5562499761581421, "learning_rate": 4.643995884038443e-06, "epoch": 0.5156701489219827, "percentage": 17.2, "elapsed_time": "1:27:39", "remaining_time": "7:01:59"}
30
+ {"current_steps": 300, "total_steps": 1686, "loss": 0.9176, "accuracy": 0.543749988079071, "learning_rate": 4.6196647777377475e-06, "epoch": 0.5334518781951545, "percentage": 17.79, "elapsed_time": "1:30:45", "remaining_time": "6:59:18"}
31
+ {"current_steps": 310, "total_steps": 1686, "loss": 0.9431, "accuracy": 0.518750011920929, "learning_rate": 4.59459764953147e-06, "epoch": 0.5512336074683263, "percentage": 18.39, "elapsed_time": "1:33:56", "remaining_time": "6:56:58"}
32
+ {"current_steps": 320, "total_steps": 1686, "loss": 0.8969, "accuracy": 0.5687500238418579, "learning_rate": 4.568803203605133e-06, "epoch": 0.5690153367414981, "percentage": 18.98, "elapsed_time": "1:36:59", "remaining_time": "6:54:02"}
33
+ {"current_steps": 330, "total_steps": 1686, "loss": 0.9255, "accuracy": 0.512499988079071, "learning_rate": 4.542290396694462e-06, "epoch": 0.58679706601467, "percentage": 19.57, "elapsed_time": "1:40:09", "remaining_time": "6:51:31"}
34
+ {"current_steps": 340, "total_steps": 1686, "loss": 0.9376, "accuracy": 0.581250011920929, "learning_rate": 4.515068434975298e-06, "epoch": 0.6045787952878418, "percentage": 20.17, "elapsed_time": "1:43:09", "remaining_time": "6:48:24"}
35
+ {"current_steps": 350, "total_steps": 1686, "loss": 0.9343, "accuracy": 0.53125, "learning_rate": 4.487146770866887e-06, "epoch": 0.6223605245610135, "percentage": 20.76, "elapsed_time": "1:46:12", "remaining_time": "6:45:25"}
36
+ {"current_steps": 360, "total_steps": 1686, "loss": 1.0312, "accuracy": 0.53125, "learning_rate": 4.458535099749666e-06, "epoch": 0.6401422538341853, "percentage": 21.35, "elapsed_time": "1:49:11", "remaining_time": "6:42:10"}
37
+ {"current_steps": 370, "total_steps": 1686, "loss": 0.9596, "accuracy": 0.550000011920929, "learning_rate": 4.429243356598694e-06, "epoch": 0.6579239831073572, "percentage": 21.95, "elapsed_time": "1:52:10", "remaining_time": "6:39:00"}
38
+ {"current_steps": 380, "total_steps": 1686, "loss": 0.8862, "accuracy": 0.512499988079071, "learning_rate": 4.399281712533875e-06, "epoch": 0.675705712380529, "percentage": 22.54, "elapsed_time": "1:55:04", "remaining_time": "6:35:29"}
39
+ {"current_steps": 390, "total_steps": 1686, "loss": 0.917, "accuracy": 0.4937500059604645, "learning_rate": 4.368660571288192e-06, "epoch": 0.6934874416537008, "percentage": 23.13, "elapsed_time": "1:58:14", "remaining_time": "6:32:54"}
40
+ {"current_steps": 400, "total_steps": 1686, "loss": 1.0041, "accuracy": 0.48750001192092896, "learning_rate": 4.337390565595163e-06, "epoch": 0.7112691709268726, "percentage": 23.72, "elapsed_time": "2:01:15", "remaining_time": "6:29:51"}
41
+ {"current_steps": 410, "total_steps": 1686, "loss": 0.868, "accuracy": 0.5687500238418579, "learning_rate": 4.305482553496786e-06, "epoch": 0.7290509002000445, "percentage": 24.32, "elapsed_time": "2:04:11", "remaining_time": "6:26:29"}
42
+ {"current_steps": 420, "total_steps": 1686, "loss": 0.9551, "accuracy": 0.5874999761581421, "learning_rate": 4.272947614573244e-06, "epoch": 0.7468326294732163, "percentage": 24.91, "elapsed_time": "2:07:06", "remaining_time": "6:23:07"}
43
+ {"current_steps": 430, "total_steps": 1686, "loss": 0.9024, "accuracy": 0.5687500238418579, "learning_rate": 4.23979704609569e-06, "epoch": 0.7646143587463881, "percentage": 25.5, "elapsed_time": "2:10:08", "remaining_time": "6:20:08"}
44
+ {"current_steps": 440, "total_steps": 1686, "loss": 0.9355, "accuracy": 0.5375000238418579, "learning_rate": 4.206042359103435e-06, "epoch": 0.78239608801956, "percentage": 26.1, "elapsed_time": "2:13:11", "remaining_time": "6:17:09"}
45
+ {"current_steps": 450, "total_steps": 1686, "loss": 0.9306, "accuracy": 0.46875, "learning_rate": 4.17169527440691e-06, "epoch": 0.8001778172927317, "percentage": 26.69, "elapsed_time": "2:16:06", "remaining_time": "6:13:50"}
46
+ {"current_steps": 460, "total_steps": 1686, "loss": 0.8402, "accuracy": 0.5625, "learning_rate": 4.136767718517797e-06, "epoch": 0.8179595465659035, "percentage": 27.28, "elapsed_time": "2:19:07", "remaining_time": "6:10:48"}
47
+ {"current_steps": 470, "total_steps": 1686, "loss": 0.9289, "accuracy": 0.4937500059604645, "learning_rate": 4.1012718195077196e-06, "epoch": 0.8357412758390753, "percentage": 27.88, "elapsed_time": "2:22:09", "remaining_time": "6:07:48"}
48
+ {"current_steps": 480, "total_steps": 1686, "loss": 0.8996, "accuracy": 0.581250011920929, "learning_rate": 4.065219902796953e-06, "epoch": 0.8535230051122472, "percentage": 28.47, "elapsed_time": "2:25:04", "remaining_time": "6:04:29"}
49
+ {"current_steps": 490, "total_steps": 1686, "loss": 0.8747, "accuracy": 0.581250011920929, "learning_rate": 4.028624486874608e-06, "epoch": 0.871304734385419, "percentage": 29.06, "elapsed_time": "2:28:05", "remaining_time": "6:01:28"}
50
+ {"current_steps": 500, "total_steps": 1686, "loss": 0.9464, "accuracy": 0.5062500238418579, "learning_rate": 3.99149827895177e-06, "epoch": 0.8890864636585908, "percentage": 29.66, "elapsed_time": "2:31:12", "remaining_time": "5:58:39"}
51
+ {"current_steps": 500, "total_steps": 1686, "eval_loss": 0.8918758630752563, "epoch": 0.8890864636585908, "percentage": 29.66, "elapsed_time": "2:37:01", "remaining_time": "6:12:27"}
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:346457aadedbdd0c4ca7d318ce433488da82e0ffc0b5d5673bac6ee8814ec494
3
+ size 5240