VERSIL91 commited on
Commit
1af2871
1 Parent(s): ff11424

End of training

Browse files
Files changed (2) hide show
  1. README.md +167 -0
  2. adapter_model.bin +3 -0
README.md ADDED
@@ -0,0 +1,167 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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: e259bfcb-1b8d-41e0-bc8b-2b5147ec7689
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
+ accelerate_config:
22
+ dynamo_backend: inductor
23
+ mixed_precision: bf16
24
+ num_machines: 1
25
+ num_processes: auto
26
+ use_cpu: false
27
+ adapter: lora
28
+ base_model: unsloth/codegemma-7b-it
29
+ bf16: auto
30
+ chat_template: llama3
31
+ dataset_prepared_path: null
32
+ datasets:
33
+ - data_files:
34
+ - 948a3386d79edd02_train_data.json
35
+ ds_type: json
36
+ format: custom
37
+ path: /workspace/input_data/948a3386d79edd02_train_data.json
38
+ type:
39
+ field_input: flores_passage
40
+ field_instruction: question
41
+ field_output: mc_answer1
42
+ format: '{instruction} {input}'
43
+ no_input_format: '{instruction}'
44
+ system_format: '{system}'
45
+ system_prompt: ''
46
+ debug: null
47
+ deepspeed: null
48
+ device_map: auto
49
+ early_stopping_patience: null
50
+ eval_max_new_tokens: 128
51
+ eval_table_size: null
52
+ evals_per_epoch: 4
53
+ flash_attention: true
54
+ fp16: null
55
+ fsdp: null
56
+ fsdp_config: null
57
+ gradient_accumulation_steps: 16
58
+ gradient_checkpointing: true
59
+ group_by_length: false
60
+ hub_model_id: VERSIL91/e259bfcb-1b8d-41e0-bc8b-2b5147ec7689
61
+ hub_repo: null
62
+ hub_strategy: checkpoint
63
+ hub_token: null
64
+ learning_rate: 0.0001
65
+ local_rank: null
66
+ logging_steps: 1
67
+ lora_alpha: 32
68
+ lora_dropout: 0.05
69
+ lora_fan_in_fan_out: null
70
+ lora_model_dir: null
71
+ lora_r: 16
72
+ lora_target_linear: true
73
+ lora_target_modules:
74
+ - q_proj
75
+ - v_proj
76
+ lr_scheduler: cosine
77
+ max_memory:
78
+ 0: 70GiB
79
+ max_steps: 50
80
+ micro_batch_size: 2
81
+ mlflow_experiment_name: /tmp/948a3386d79edd02_train_data.json
82
+ model_type: AutoModelForCausalLM
83
+ num_epochs: 1
84
+ optimizer: adamw_bnb_8bit
85
+ output_dir: miner_id_24
86
+ pad_to_sequence_len: true
87
+ quantization_config:
88
+ llm_int8_enable_fp32_cpu_offload: true
89
+ load_in_8bit: true
90
+ resume_from_checkpoint: null
91
+ s2_attention: null
92
+ sample_packing: false
93
+ saves_per_epoch: 4
94
+ sequence_len: 4056
95
+ strict: false
96
+ tf32: false
97
+ tokenizer_type: AutoTokenizer
98
+ torch_compile: true
99
+ train_on_inputs: false
100
+ trust_remote_code: true
101
+ val_set_size: 0.05
102
+ wandb_entity: null
103
+ wandb_mode: online
104
+ wandb_name: e259bfcb-1b8d-41e0-bc8b-2b5147ec7689
105
+ wandb_project: Gradients-On-Demand
106
+ wandb_run: your_name
107
+ wandb_runid: e259bfcb-1b8d-41e0-bc8b-2b5147ec7689
108
+ warmup_steps: 10
109
+ weight_decay: 0.0
110
+ xformers_attention: null
111
+
112
+ ```
113
+
114
+ </details><br>
115
+
116
+ # e259bfcb-1b8d-41e0-bc8b-2b5147ec7689
117
+
118
+ This model is a fine-tuned version of [unsloth/codegemma-7b-it](https://huggingface.co/unsloth/codegemma-7b-it) on the None dataset.
119
+ It achieves the following results on the evaluation set:
120
+ - Loss: 2.5298
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: 0.0001
140
+ - train_batch_size: 2
141
+ - eval_batch_size: 2
142
+ - seed: 42
143
+ - gradient_accumulation_steps: 16
144
+ - total_train_batch_size: 32
145
+ - optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
146
+ - lr_scheduler_type: cosine
147
+ - lr_scheduler_warmup_steps: 10
148
+ - training_steps: 16
149
+
150
+ ### Training results
151
+
152
+ | Training Loss | Epoch | Step | Validation Loss |
153
+ |:-------------:|:------:|:----:|:---------------:|
154
+ | 4.515 | 0.0661 | 1 | 4.7045 |
155
+ | 4.5637 | 0.2645 | 4 | 4.3521 |
156
+ | 3.4257 | 0.5289 | 8 | 3.0778 |
157
+ | 3.1827 | 0.7934 | 12 | 2.6660 |
158
+ | 4.6603 | 1.0579 | 16 | 2.5298 |
159
+
160
+
161
+ ### Framework versions
162
+
163
+ - PEFT 0.13.2
164
+ - Transformers 4.46.0
165
+ - Pytorch 2.5.0+cu124
166
+ - Datasets 3.0.1
167
+ - Tokenizers 0.20.1
adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37633cfa9063a3d39f800ee4d3a39462bbdf0d08ed5903f51aabc54970b9b4b5
3
+ size 200157226