Hugging Face's logo

hon9kon9ize
/
cantonese-gemma-sft-lora

PEFT
TensorBoard
Safetensors
gemma
alignment-handbook
trl
sft
Generated from Trainer
Model card Files Metrics Community
indiejoseph commited on
Commit
95a3e56
·
verified ·
1 Parent(s): e5ca3b4

Model save

Browse files
Files changed (11) hide show
  1. README.md +8 -11
  2. adapter_config.json +10 -7
  3. adapter_model.safetensors +2 -2
  4. all_results.json +5 -5
  5. runs/Apr20_16-05-37_1f2756113ce9/events.out.tfevents.1713629403.1f2756113ce9.9459.0 +3 -0
  6. runs/Apr20_16-12-10_1f2756113ce9/events.out.tfevents.1713629570.1f2756113ce9.13142.0 +3 -0
  7. runs/Apr20_16-14-55_1f2756113ce9/events.out.tfevents.1713629895.1f2756113ce9.14711.0 +3 -0
  8. runs/Apr20_16-19-11_1f2756113ce9/events.out.tfevents.1713629983.1f2756113ce9.17085.0 +3 -0
  9. train_results.json +5 -5
  10. trainer_state.json +0 -0
  11. training_args.bin +1 -1
README.md CHANGED
@@ -2,16 +2,13 @@
2
  license: gemma
3
  library_name: peft
4
  tags:
5
- - alignment-handbook
6
- - trl
7
- - sft
8
- - generated_from_trainer
9
  - trl
10
  - sft
 
11
  - generated_from_trainer
12
  base_model: google/gemma-2b
13
  datasets:
14
- - HuggingFaceH4/deita-10k-v0-sft
15
  model-index:
16
  - name: cantonese-gemma-sft-lora
17
  results: []
@@ -22,7 +19,7 @@ should probably proofread and complete it, then remove this comment. -->
22
 
23
  # cantonese-gemma-sft-lora
24
 
25
- This model is a fine-tuned version of [google/gemma-2b](https://huggingface.co/google/gemma-2b) on the HuggingFaceH4/deita-10k-v0-sft dataset.
26
 
27
  ## Model description
28
 
@@ -41,16 +38,16 @@ More information needed
41
  ### Training hyperparameters
42
 
43
  The following hyperparameters were used during training:
44
- - learning_rate: 0.0002
45
- - train_batch_size: 2
46
- - eval_batch_size: 4
47
  - seed: 42
48
  - distributed_type: multi-GPU
49
  - gradient_accumulation_steps: 4
50
- - total_train_batch_size: 8
51
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
52
  - lr_scheduler_type: cosine
53
- - lr_scheduler_warmup_ratio: 0.1
54
  - num_epochs: 1
55
 
56
  ### Training results
 
2
  license: gemma
3
  library_name: peft
4
  tags:
 
 
 
 
5
  - trl
6
  - sft
7
+ - alignment-handbook
8
  - generated_from_trainer
9
  base_model: google/gemma-2b
10
  datasets:
11
+ - generator
12
  model-index:
13
  - name: cantonese-gemma-sft-lora
14
  results: []
 
19
 
20
  # cantonese-gemma-sft-lora
21
 
22
+ This model is a fine-tuned version of [google/gemma-2b](https://huggingface.co/google/gemma-2b) on the generator dataset.
23
 
24
  ## Model description
25
 
 
38
  ### Training hyperparameters
39
 
40
  The following hyperparameters were used during training:
41
+ - learning_rate: 0.0001
42
+ - train_batch_size: 1
43
+ - eval_batch_size: 2
44
  - seed: 42
45
  - distributed_type: multi-GPU
46
  - gradient_accumulation_steps: 4
47
+ - total_train_batch_size: 4
48
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
49
  - lr_scheduler_type: cosine
50
+ - lr_scheduler_warmup_ratio: 0.03
51
  - num_epochs: 1
52
 
53
  ### Training results
adapter_config.json CHANGED
@@ -9,23 +9,26 @@
9
  "layers_pattern": null,
10
  "layers_to_transform": null,
11
  "loftq_config": {},
12
- "lora_alpha": 256,
13
  "lora_dropout": 0.05,
14
  "megatron_config": null,
15
  "megatron_core": "megatron.core",
16
- "modules_to_save": null,
 
 
 
17
  "peft_type": "LORA",
18
- "r": 128,
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
22
- "o_proj",
23
  "down_proj",
24
- "up_proj",
25
- "k_proj",
26
  "v_proj",
 
27
  "q_proj",
28
- "gate_proj"
 
 
29
  ],
30
  "task_type": "CAUSAL_LM"
31
  }
 
9
  "layers_pattern": null,
10
  "layers_to_transform": null,
11
  "loftq_config": {},
12
+ "lora_alpha": 16,
13
  "lora_dropout": 0.05,
14
  "megatron_config": null,
15
  "megatron_core": "megatron.core",
16
+ "modules_to_save": [
17
+ "embed_tokens",
18
+ "lm_head"
19
+ ],
20
  "peft_type": "LORA",
21
+ "r": 8,
22
  "rank_pattern": {},
23
  "revision": null,
24
  "target_modules": [
 
25
  "down_proj",
 
 
26
  "v_proj",
27
+ "gate_proj",
28
  "q_proj",
29
+ "k_proj",
30
+ "o_proj",
31
+ "up_proj"
32
  ],
33
  "task_type": "CAUSAL_LM"
34
  }
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8c7923dfbbdf22a477528175cf5f0fce7c03d51b33c5f50d02292859f7eb1cc5
3
- size 313820752
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02c5fcf66f6176c69a169349d93b39b72262efc26babaf7c23d376c31751f6e6
3
+ size 2116798424
all_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 1.0,
3
- "train_loss": 1.0868773834192678,
4
- "train_runtime": 33367.2438,
5
- "train_samples": 9500,
6
- "train_samples_per_second": 1.149,
7
- "train_steps_per_second": 0.144
8
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "train_loss": 1.0705042276791343,
4
+ "train_runtime": 37728.5727,
5
+ "train_samples": 26282,
6
+ "train_samples_per_second": 1.041,
7
+ "train_steps_per_second": 0.26
8
  }
runs/Apr20_16-05-37_1f2756113ce9/events.out.tfevents.1713629403.1f2756113ce9.9459.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4b225024939427a42e751934002308229a4e628c8e90810fbd40af8bee2d080
3
+ size 4910
runs/Apr20_16-12-10_1f2756113ce9/events.out.tfevents.1713629570.1f2756113ce9.13142.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33beea90171217f79d94c4030c2ffafd8c7e132ab2e62f28c75a7421b0e15e02
3
+ size 5945
runs/Apr20_16-14-55_1f2756113ce9/events.out.tfevents.1713629895.1f2756113ce9.14711.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f27015d821f1edccb3742565314556088bb98287050ead212bb7c9fd2720e054
3
+ size 4910
runs/Apr20_16-19-11_1f2756113ce9/events.out.tfevents.1713629983.1f2756113ce9.17085.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:290de46f53f2ed1b1311e0d7a24e92b1e04f86b952550218802a4f5b643f96db
3
+ size 419564
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 1.0,
3
- "train_loss": 1.0868773834192678,
4
- "train_runtime": 33367.2438,
5
- "train_samples": 9500,
6
- "train_samples_per_second": 1.149,
7
- "train_steps_per_second": 0.144
8
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "train_loss": 1.0705042276791343,
4
+ "train_runtime": 37728.5727,
5
+ "train_samples": 26282,
6
+ "train_samples_per_second": 1.041,
7
+ "train_steps_per_second": 0.26
8
  }
trainer_state.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a5ddf2fa92c9d7a52102e7ca0b2be8f02f3d748b873ff07106f05cedecebdf3d
3
  size 5112
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49e66a6c2dc2f611c5a6f0f0f18c17b26498dc36d58718b053cd79aef21bd000
3
  size 5112