diaenra commited on
Commit
63ff89d
1 Parent(s): fb93267

End of training

Browse files
Files changed (2) hide show
  1. README.md +155 -0
  2. adapter_model.bin +3 -0
README.md ADDED
@@ -0,0 +1,155 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ license: llama3
4
+ base_model: tokyotech-llm/Llama-3-Swallow-8B-v0.1
5
+ tags:
6
+ - axolotl
7
+ - generated_from_trainer
8
+ model-index:
9
+ - name: 7c697e4a-2919-4f0a-8545-ce302d164ccc
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: tokyotech-llm/Llama-3-Swallow-8B-v0.1
23
+ bf16: auto
24
+ chat_template: llama3
25
+ dataset_prepared_path: null
26
+ datasets:
27
+ - data_files:
28
+ - ebcf23a45eabafb9_train_data.json
29
+ ds_type: json
30
+ format: custom
31
+ path: /workspace/input_data/ebcf23a45eabafb9_train_data.json
32
+ type:
33
+ field_input: context_processed
34
+ field_instruction: processed
35
+ field_output: raw
36
+ format: '{instruction} {input}'
37
+ no_input_format: '{instruction}'
38
+ system_format: '{system}'
39
+ system_prompt: ''
40
+ debug: null
41
+ deepspeed: null
42
+ early_stopping_patience: 1
43
+ eval_max_new_tokens: 128
44
+ eval_steps: 25
45
+ eval_table_size: null
46
+ flash_attention: true
47
+ fp16: false
48
+ fsdp: null
49
+ fsdp_config: null
50
+ gradient_accumulation_steps: 4
51
+ gradient_checkpointing: true
52
+ group_by_length: true
53
+ hub_model_id: diaenra/7c697e4a-2919-4f0a-8545-ce302d164ccc
54
+ hub_repo: null
55
+ hub_strategy: checkpoint
56
+ hub_token: null
57
+ learning_rate: 0.0002
58
+ load_in_4bit: false
59
+ load_in_8bit: false
60
+ local_rank: null
61
+ logging_steps: 1
62
+ lora_alpha: 32
63
+ lora_dropout: 0.05
64
+ lora_fan_in_fan_out: null
65
+ lora_model_dir: null
66
+ lora_r: 16
67
+ lora_target_linear: true
68
+ lr_scheduler: cosine
69
+ max_steps: 100
70
+ micro_batch_size: 2
71
+ mlflow_experiment_name: /tmp/ebcf23a45eabafb9_train_data.json
72
+ model_type: AutoModelForCausalLM
73
+ num_epochs: 1
74
+ optimizer: adamw_torch
75
+ output_dir: miner_id_24
76
+ pad_to_sequence_len: true
77
+ resume_from_checkpoint: null
78
+ s2_attention: null
79
+ sample_packing: false
80
+ save_steps: 25
81
+ sequence_len: 4056
82
+ special_tokens:
83
+ pad_token: <|end_of_text|>
84
+ strict: false
85
+ tf32: false
86
+ tokenizer_type: AutoTokenizer
87
+ train_on_inputs: false
88
+ trust_remote_code: true
89
+ val_set_size: 0.05
90
+ wandb_entity: diaenra-tao-miner
91
+ wandb_mode: online
92
+ wandb_name: 7c697e4a-2919-4f0a-8545-ce302d164ccc
93
+ wandb_project: tao
94
+ wandb_run: diaenra
95
+ wandb_runid: 7c697e4a-2919-4f0a-8545-ce302d164ccc
96
+ warmup_ratio: 0.05
97
+ weight_decay: 0.01
98
+ xformers_attention: true
99
+
100
+ ```
101
+
102
+ </details><br>
103
+
104
+ # 7c697e4a-2919-4f0a-8545-ce302d164ccc
105
+
106
+ This model is a fine-tuned version of [tokyotech-llm/Llama-3-Swallow-8B-v0.1](https://huggingface.co/tokyotech-llm/Llama-3-Swallow-8B-v0.1) on the None dataset.
107
+ It achieves the following results on the evaluation set:
108
+ - Loss: 0.3504
109
+
110
+ ## Model description
111
+
112
+ More information needed
113
+
114
+ ## Intended uses & limitations
115
+
116
+ More information needed
117
+
118
+ ## Training and evaluation data
119
+
120
+ More information needed
121
+
122
+ ## Training procedure
123
+
124
+ ### Training hyperparameters
125
+
126
+ The following hyperparameters were used during training:
127
+ - learning_rate: 0.0002
128
+ - train_batch_size: 2
129
+ - eval_batch_size: 2
130
+ - seed: 42
131
+ - gradient_accumulation_steps: 4
132
+ - total_train_batch_size: 8
133
+ - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
134
+ - lr_scheduler_type: cosine
135
+ - lr_scheduler_warmup_steps: 5
136
+ - training_steps: 100
137
+
138
+ ### Training results
139
+
140
+ | Training Loss | Epoch | Step | Validation Loss |
141
+ |:-------------:|:------:|:----:|:---------------:|
142
+ | 0.4435 | 0.0000 | 1 | 1.1717 |
143
+ | 0.3282 | 0.0005 | 25 | 0.3831 |
144
+ | 0.6431 | 0.0010 | 50 | 0.3806 |
145
+ | 0.2269 | 0.0015 | 75 | 0.3507 |
146
+ | 0.4125 | 0.0020 | 100 | 0.3504 |
147
+
148
+
149
+ ### Framework versions
150
+
151
+ - PEFT 0.13.2
152
+ - Transformers 4.46.0
153
+ - Pytorch 2.5.0+cu124
154
+ - Datasets 3.0.1
155
+ - Tokenizers 0.20.1
adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6adfa0737a0b90e0ae790f6a8b4be857e5a9df1d9dc9bf3c9f6b831b4f650fe9
3
+ size 167934026