dada22231 commited on
Commit
6dd2469
1 Parent(s): 0bb4978

End of training

Browse files
Files changed (2) hide show
  1. README.md +172 -0
  2. adapter_model.bin +3 -0
README.md ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ license: apache-2.0
4
+ base_model: unsloth/codegemma-7b-it
5
+ tags:
6
+ - axolotl
7
+ - generated_from_trainer
8
+ model-index:
9
+ - name: 0ad7d4ad-f79e-490c-8ae1-ca2db3df3dcc
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/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
17
+ <details><summary>See axolotl config</summary>
18
+
19
+ axolotl version: `0.4.1`
20
+ ```yaml
21
+ adapter: lora
22
+ base_model: unsloth/codegemma-7b-it
23
+ bf16: auto
24
+ chat_template: llama3
25
+ cosine_min_lr_ratio: 0.1
26
+ data_processes: 16
27
+ dataset_prepared_path: null
28
+ datasets:
29
+ - data_files:
30
+ - 7b678bff7406b34b_train_data.json
31
+ ds_type: json
32
+ format: custom
33
+ path: /workspace/input_data/7b678bff7406b34b_train_data.json
34
+ type:
35
+ field_instruction: instruction
36
+ field_output: output
37
+ format: '{instruction}'
38
+ no_input_format: '{instruction}'
39
+ system_format: '{system}'
40
+ system_prompt: ''
41
+ debug: null
42
+ deepspeed: null
43
+ device_map: auto
44
+ do_eval: true
45
+ eval_batch_size: 1
46
+ eval_sample_packing: false
47
+ eval_steps: 25
48
+ evaluation_strategy: steps
49
+ flash_attention: false
50
+ fp16: null
51
+ fsdp: null
52
+ fsdp_config: null
53
+ gradient_accumulation_steps: 32
54
+ gradient_checkpointing: true
55
+ group_by_length: true
56
+ hub_model_id: dada22231/0ad7d4ad-f79e-490c-8ae1-ca2db3df3dcc
57
+ hub_strategy: checkpoint
58
+ hub_token: null
59
+ hub_username: dada22231
60
+ learning_rate: 0.0001
61
+ load_in_4bit: false
62
+ load_in_8bit: false
63
+ local_rank: null
64
+ logging_steps: 1
65
+ lora_alpha: 64
66
+ lora_dropout: 0.05
67
+ lora_fan_in_fan_out: null
68
+ lora_model_dir: null
69
+ lora_r: 32
70
+ lora_target_linear: true
71
+ lora_target_modules:
72
+ - q_proj
73
+ - v_proj
74
+ lr_scheduler: cosine
75
+ max_grad_norm: 1.0
76
+ max_memory:
77
+ 0: 70GiB
78
+ 1: 70GiB
79
+ 2: 70GiB
80
+ 3: 70GiB
81
+ max_steps: 50
82
+ micro_batch_size: 1
83
+ mlflow_experiment_name: /tmp/7b678bff7406b34b_train_data.json
84
+ model_type: AutoModelForCausalLM
85
+ num_epochs: 3
86
+ optim_args:
87
+ adam_beta1: 0.9
88
+ adam_beta2: 0.95
89
+ adam_epsilon: 1e-5
90
+ optimizer: adamw_torch
91
+ output_dir: miner_id_24
92
+ pad_to_sequence_len: true
93
+ repository_id: dada22231/0ad7d4ad-f79e-490c-8ae1-ca2db3df3dcc
94
+ resume_from_checkpoint: null
95
+ s2_attention: null
96
+ sample_packing: false
97
+ save_steps: 25
98
+ save_strategy: steps
99
+ sequence_len: 2048
100
+ strict: false
101
+ tf32: false
102
+ tokenizer_type: AutoTokenizer
103
+ torch_compile: false
104
+ train_on_inputs: false
105
+ trust_remote_code: true
106
+ val_set_size: 50
107
+ wandb_entity: null
108
+ wandb_mode: online
109
+ wandb_name: 0ad7d4ad-f79e-490c-8ae1-ca2db3df3dcc
110
+ wandb_project: Public_TuningSN
111
+ wandb_runid: 0ad7d4ad-f79e-490c-8ae1-ca2db3df3dcc
112
+ warmup_ratio: 0.04
113
+ weight_decay: 0.01
114
+ xformers_attention: null
115
+
116
+ ```
117
+
118
+ </details><br>
119
+
120
+ # 0ad7d4ad-f79e-490c-8ae1-ca2db3df3dcc
121
+
122
+ This model is a fine-tuned version of [unsloth/codegemma-7b-it](https://huggingface.co/unsloth/codegemma-7b-it) on the None dataset.
123
+ It achieves the following results on the evaluation set:
124
+ - Loss: 0.9967
125
+
126
+ ## Model description
127
+
128
+ More information needed
129
+
130
+ ## Intended uses & limitations
131
+
132
+ More information needed
133
+
134
+ ## Training and evaluation data
135
+
136
+ More information needed
137
+
138
+ ## Training procedure
139
+
140
+ ### Training hyperparameters
141
+
142
+ The following hyperparameters were used during training:
143
+ - learning_rate: 0.0001
144
+ - train_batch_size: 1
145
+ - eval_batch_size: 1
146
+ - seed: 42
147
+ - distributed_type: multi-GPU
148
+ - num_devices: 4
149
+ - gradient_accumulation_steps: 32
150
+ - total_train_batch_size: 128
151
+ - total_eval_batch_size: 4
152
+ - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=adam_beta1=0.9,adam_beta2=0.95,adam_epsilon=1e-5
153
+ - lr_scheduler_type: cosine
154
+ - lr_scheduler_warmup_steps: 2
155
+ - training_steps: 50
156
+
157
+ ### Training results
158
+
159
+ | Training Loss | Epoch | Step | Validation Loss |
160
+ |:-------------:|:------:|:----:|:---------------:|
161
+ | 1.0758 | 0.0137 | 1 | 2.0613 |
162
+ | 1.0646 | 0.3428 | 25 | 1.0579 |
163
+ | 0.8543 | 0.6855 | 50 | 0.9967 |
164
+
165
+
166
+ ### Framework versions
167
+
168
+ - PEFT 0.13.2
169
+ - Transformers 4.46.0
170
+ - Pytorch 2.5.0+cu124
171
+ - Datasets 3.0.1
172
+ - Tokenizers 0.20.1
adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37c42795f432537821d0b3c0b6cd39b71037f85fda072a5299595439f13fdf29
3
+ size 400173482