Hugging Face's logo

PhoenixB
/
c6460c46-b09a-40f9-8e48-e0bc0d299c2e

PEFT
Safetensors
qwen2
axolotl
Generated from Trainer
Model card Files Community
1

Improve language tag

#1
by lbourdois - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+210
-196
Files changed (1) hide show
  1. README.md +210 -196
README.md CHANGED
@@ -1,197 +1,211 @@
1
- ---
2
- library_name: peft
3
- license: apache-2.0
4
- base_model: Qwen/Qwen2.5-7B
5
- tags:
6
- - axolotl
7
- - generated_from_trainer
8
- model-index:
9
- - name: c6460c46-b09a-40f9-8e48-e0bc0d299c2e
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: `0.5.2`
20
- ```yaml
21
- adapter: lora
22
- auto_find_batch_size: true
23
- base_model: Qwen/Qwen2.5-7B
24
- bf16: auto
25
- chat_template: llama3
26
- dataset_prepared_path: null
27
- datasets:
28
- - data_files:
29
- - e67a3bba658eaa0f_train_data.json
30
- ds_type: json
31
- format: custom
32
- path: /workspace/input_data/e67a3bba658eaa0f_train_data.json
33
- type:
34
- field_input: input
35
- field_instruction: system_prompt
36
- field_output: reference_answer
37
- format: '{instruction} {input}'
38
- no_input_format: '{instruction}'
39
- system_format: '{system}'
40
- system_prompt: ''
41
- debug: null
42
- deepspeed: /workspace/axolotl/configs/deepspeed_stage2.json
43
- early_stopping_patience: 5
44
- eval_max_new_tokens: 128
45
- eval_sample_packing: false
46
- eval_steps: 50
47
- eval_table_size: null
48
- evaluation_strategy: steps
49
- flash_attention: true
50
- fp16: false
51
- gpu_memory_limit: 80GiB
52
- gradient_accumulation_steps: 2
53
- gradient_checkpointing: true
54
- greater_is_better: false
55
- group_by_length: true
56
- hub_model_id: PhoenixB/c6460c46-b09a-40f9-8e48-e0bc0d299c2e
57
- hub_repo: null
58
- hub_strategy: checkpoint
59
- hub_token: null
60
- learning_rate: 2e-4
61
- liger_fused_linear_cross_entropy: true
62
- liger_glu_activation: true
63
- liger_layer_norm: true
64
- liger_rms_norm: true
65
- liger_rope: true
66
- load_best_model_at_end: true
67
- load_in_4bit: false
68
- load_in_8bit: false
69
- local_rank: null
70
- logging_steps: 5
71
- lora_alpha: 16
72
- lora_dropout: 0.05
73
- lora_fan_in_fan_out: null
74
- lora_model_dir: null
75
- lora_r: 16
76
- lora_target_linear: true
77
- lr_scheduler: cosine
78
- max_steps: 10000
79
- metric_for_best_model: eval_loss
80
- micro_batch_size: 2
81
- mlflow_experiment_name: /tmp/e67a3bba658eaa0f_train_data.json
82
- model_type: AutoModelForCausalLM
83
- num_epochs: 3
84
- optimizer: adamw_torch_fused
85
- output_dir: miner_id_24
86
- pad_to_sequence_len: true
87
- plugins:
88
- - axolotl.integrations.liger.LigerPlugin
89
- resume_from_checkpoint: null
90
- s2_attention: null
91
- sample_packing: false
92
- save_steps: 50
93
- save_total_limit: 1
94
- sequence_len: 8196
95
- strict: false
96
- tf32: true
97
- tokenizer_type: AutoTokenizer
98
- train_on_inputs: false
99
- trust_remote_code: true
100
- val_set_size: 0.05
101
- wandb_entity: null
102
- wandb_mode: online
103
- wandb_name: c4b7f349-24bc-4e90-8602-b3a61dc8b8ab
104
- wandb_project: Gradients-On-Demand
105
- wandb_run: your_name
106
- wandb_runid: c4b7f349-24bc-4e90-8602-b3a61dc8b8ab
107
- warmup_steps: 20
108
- weight_decay: 0.0
109
-
110
- ```
111
-
112
- </details><br>
113
-
114
- # c6460c46-b09a-40f9-8e48-e0bc0d299c2e
115
-
116
- This model is a fine-tuned version of [Qwen/Qwen2.5-7B](https://huggingface.co/Qwen/Qwen2.5-7B) on the None dataset.
117
- It achieves the following results on the evaluation set:
118
- - Loss: 0.0151
119
-
120
- ## Model description
121
-
122
- More information needed
123
-
124
- ## Intended uses & limitations
125
-
126
- More information needed
127
-
128
- ## Training and evaluation data
129
-
130
- More information needed
131
-
132
- ## Training procedure
133
-
134
- ### Training hyperparameters
135
-
136
- The following hyperparameters were used during training:
137
- - learning_rate: 0.0002
138
- - train_batch_size: 2
139
- - eval_batch_size: 2
140
- - seed: 42
141
- - distributed_type: multi-GPU
142
- - num_devices: 2
143
- - gradient_accumulation_steps: 2
144
- - total_train_batch_size: 8
145
- - total_eval_batch_size: 4
146
- - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
147
- - lr_scheduler_type: cosine
148
- - lr_scheduler_warmup_steps: 20
149
- - training_steps: 10000
150
-
151
- ### Training results
152
-
153
- | Training Loss | Epoch | Step | Validation Loss |
154
- |:-------------:|:------:|:----:|:---------------:|
155
- | No log | 0.0001 | 1 | 1.3133 |
156
- | 1.3321 | 0.0056 | 50 | 1.1406 |
157
- | 1.095 | 0.0111 | 100 | 0.9931 |
158
- | 0.7319 | 0.0167 | 150 | 0.8583 |
159
- | 0.4488 | 0.0223 | 200 | 0.7062 |
160
- | 0.5606 | 0.0279 | 250 | 0.6025 |
161
- | 0.369 | 0.0334 | 300 | 0.4909 |
162
- | 0.3071 | 0.0390 | 350 | 0.4296 |
163
- | 0.1691 | 0.0446 | 400 | 0.3548 |
164
- | 0.1887 | 0.0502 | 450 | 0.2855 |
165
- | 0.2134 | 0.0557 | 500 | 0.2420 |
166
- | 0.0553 | 0.0613 | 550 | 0.2010 |
167
- | 0.105 | 0.0669 | 600 | 0.1763 |
168
- | 0.0427 | 0.0724 | 650 | 0.1375 |
169
- | 0.0575 | 0.0780 | 700 | 0.1137 |
170
- | 0.1403 | 0.0836 | 750 | 0.0975 |
171
- | 0.0601 | 0.0892 | 800 | 0.0804 |
172
- | 0.0619 | 0.0947 | 850 | 0.0610 |
173
- | 0.0524 | 0.1003 | 900 | 0.0500 |
174
- | 0.1153 | 0.1059 | 950 | 0.0399 |
175
- | 0.0138 | 0.1115 | 1000 | 0.0333 |
176
- | 0.0136 | 0.1170 | 1050 | 0.0298 |
177
- | 0.01 | 0.1226 | 1100 | 0.0271 |
178
- | 0.0228 | 0.1282 | 1150 | 0.0201 |
179
- | 0.0364 | 0.1337 | 1200 | 0.0175 |
180
- | 0.0149 | 0.1393 | 1250 | 0.0167 |
181
- | 0.0368 | 0.1449 | 1300 | 0.0184 |
182
- | 0.0135 | 0.1505 | 1350 | 0.0151 |
183
- | 0.0064 | 0.1560 | 1400 | 0.0133 |
184
- | 0.0332 | 0.1616 | 1450 | 0.0144 |
185
- | 0.009 | 0.1672 | 1500 | 0.0145 |
186
- | 0.0073 | 0.1728 | 1550 | 0.0137 |
187
- | 0.0122 | 0.1783 | 1600 | 0.0148 |
188
- | 0.0057 | 0.1839 | 1650 | 0.0151 |
189
-
190
-
191
- ### Framework versions
192
-
193
- - PEFT 0.13.2
194
- - Transformers 4.46.3
195
- - Pytorch 2.5.1+cu124
196
- - Datasets 3.1.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
197
  - Tokenizers 0.20.3
 
1
+ ---
2
+ library_name: peft
3
+ license: apache-2.0
4
+ base_model: Qwen/Qwen2.5-7B
5
+ tags:
6
+ - axolotl
7
+ - generated_from_trainer
8
+ language:
9
+ - zho
10
+ - eng
11
+ - fra
12
+ - spa
13
+ - por
14
+ - deu
15
+ - ita
16
+ - rus
17
+ - jpn
18
+ - kor
19
+ - vie
20
+ - tha
21
+ - ara
22
+ model-index:
23
+ - name: c6460c46-b09a-40f9-8e48-e0bc0d299c2e
24
+ results: []
25
+ ---
26
+
27
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
28
+ should probably proofread and complete it, then remove this comment. -->
29
+
30
+ [<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)
31
+ <details><summary>See axolotl config</summary>
32
+
33
+ axolotl version: `0.5.2`
34
+ ```yaml
35
+ adapter: lora
36
+ auto_find_batch_size: true
37
+ base_model: Qwen/Qwen2.5-7B
38
+ bf16: auto
39
+ chat_template: llama3
40
+ dataset_prepared_path: null
41
+ datasets:
42
+ - data_files:
43
+ - e67a3bba658eaa0f_train_data.json
44
+ ds_type: json
45
+ format: custom
46
+ path: /workspace/input_data/e67a3bba658eaa0f_train_data.json
47
+ type:
48
+ field_input: input
49
+ field_instruction: system_prompt
50
+ field_output: reference_answer
51
+ format: '{instruction} {input}'
52
+ no_input_format: '{instruction}'
53
+ system_format: '{system}'
54
+ system_prompt: ''
55
+ debug: null
56
+ deepspeed: /workspace/axolotl/configs/deepspeed_stage2.json
57
+ early_stopping_patience: 5
58
+ eval_max_new_tokens: 128
59
+ eval_sample_packing: false
60
+ eval_steps: 50
61
+ eval_table_size: null
62
+ evaluation_strategy: steps
63
+ flash_attention: true
64
+ fp16: false
65
+ gpu_memory_limit: 80GiB
66
+ gradient_accumulation_steps: 2
67
+ gradient_checkpointing: true
68
+ greater_is_better: false
69
+ group_by_length: true
70
+ hub_model_id: PhoenixB/c6460c46-b09a-40f9-8e48-e0bc0d299c2e
71
+ hub_repo: null
72
+ hub_strategy: checkpoint
73
+ hub_token: null
74
+ learning_rate: 2e-4
75
+ liger_fused_linear_cross_entropy: true
76
+ liger_glu_activation: true
77
+ liger_layer_norm: true
78
+ liger_rms_norm: true
79
+ liger_rope: true
80
+ load_best_model_at_end: true
81
+ load_in_4bit: false
82
+ load_in_8bit: false
83
+ local_rank: null
84
+ logging_steps: 5
85
+ lora_alpha: 16
86
+ lora_dropout: 0.05
87
+ lora_fan_in_fan_out: null
88
+ lora_model_dir: null
89
+ lora_r: 16
90
+ lora_target_linear: true
91
+ lr_scheduler: cosine
92
+ max_steps: 10000
93
+ metric_for_best_model: eval_loss
94
+ micro_batch_size: 2
95
+ mlflow_experiment_name: /tmp/e67a3bba658eaa0f_train_data.json
96
+ model_type: AutoModelForCausalLM
97
+ num_epochs: 3
98
+ optimizer: adamw_torch_fused
99
+ output_dir: miner_id_24
100
+ pad_to_sequence_len: true
101
+ plugins:
102
+ - axolotl.integrations.liger.LigerPlugin
103
+ resume_from_checkpoint: null
104
+ s2_attention: null
105
+ sample_packing: false
106
+ save_steps: 50
107
+ save_total_limit: 1
108
+ sequence_len: 8196
109
+ strict: false
110
+ tf32: true
111
+ tokenizer_type: AutoTokenizer
112
+ train_on_inputs: false
113
+ trust_remote_code: true
114
+ val_set_size: 0.05
115
+ wandb_entity: null
116
+ wandb_mode: online
117
+ wandb_name: c4b7f349-24bc-4e90-8602-b3a61dc8b8ab
118
+ wandb_project: Gradients-On-Demand
119
+ wandb_run: your_name
120
+ wandb_runid: c4b7f349-24bc-4e90-8602-b3a61dc8b8ab
121
+ warmup_steps: 20
122
+ weight_decay: 0.0
123
+
124
+ ```
125
+
126
+ </details><br>
127
+
128
+ # c6460c46-b09a-40f9-8e48-e0bc0d299c2e
129
+
130
+ This model is a fine-tuned version of [Qwen/Qwen2.5-7B](https://huggingface.co/Qwen/Qwen2.5-7B) on the None dataset.
131
+ It achieves the following results on the evaluation set:
132
+ - Loss: 0.0151
133
+
134
+ ## Model description
135
+
136
+ More information needed
137
+
138
+ ## Intended uses & limitations
139
+
140
+ More information needed
141
+
142
+ ## Training and evaluation data
143
+
144
+ More information needed
145
+
146
+ ## Training procedure
147
+
148
+ ### Training hyperparameters
149
+
150
+ The following hyperparameters were used during training:
151
+ - learning_rate: 0.0002
152
+ - train_batch_size: 2
153
+ - eval_batch_size: 2
154
+ - seed: 42
155
+ - distributed_type: multi-GPU
156
+ - num_devices: 2
157
+ - gradient_accumulation_steps: 2
158
+ - total_train_batch_size: 8
159
+ - total_eval_batch_size: 4
160
+ - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
161
+ - lr_scheduler_type: cosine
162
+ - lr_scheduler_warmup_steps: 20
163
+ - training_steps: 10000
164
+
165
+ ### Training results
166
+
167
+ | Training Loss | Epoch | Step | Validation Loss |
168
+ |:-------------:|:------:|:----:|:---------------:|
169
+ | No log | 0.0001 | 1 | 1.3133 |
170
+ | 1.3321 | 0.0056 | 50 | 1.1406 |
171
+ | 1.095 | 0.0111 | 100 | 0.9931 |
172
+ | 0.7319 | 0.0167 | 150 | 0.8583 |
173
+ | 0.4488 | 0.0223 | 200 | 0.7062 |
174
+ | 0.5606 | 0.0279 | 250 | 0.6025 |
175
+ | 0.369 | 0.0334 | 300 | 0.4909 |
176
+ | 0.3071 | 0.0390 | 350 | 0.4296 |
177
+ | 0.1691 | 0.0446 | 400 | 0.3548 |
178
+ | 0.1887 | 0.0502 | 450 | 0.2855 |
179
+ | 0.2134 | 0.0557 | 500 | 0.2420 |
180
+ | 0.0553 | 0.0613 | 550 | 0.2010 |
181
+ | 0.105 | 0.0669 | 600 | 0.1763 |
182
+ | 0.0427 | 0.0724 | 650 | 0.1375 |
183
+ | 0.0575 | 0.0780 | 700 | 0.1137 |
184
+ | 0.1403 | 0.0836 | 750 | 0.0975 |
185
+ | 0.0601 | 0.0892 | 800 | 0.0804 |
186
+ | 0.0619 | 0.0947 | 850 | 0.0610 |
187
+ | 0.0524 | 0.1003 | 900 | 0.0500 |
188
+ | 0.1153 | 0.1059 | 950 | 0.0399 |
189
+ | 0.0138 | 0.1115 | 1000 | 0.0333 |
190
+ | 0.0136 | 0.1170 | 1050 | 0.0298 |
191
+ | 0.01 | 0.1226 | 1100 | 0.0271 |
192
+ | 0.0228 | 0.1282 | 1150 | 0.0201 |
193
+ | 0.0364 | 0.1337 | 1200 | 0.0175 |
194
+ | 0.0149 | 0.1393 | 1250 | 0.0167 |
195
+ | 0.0368 | 0.1449 | 1300 | 0.0184 |
196
+ | 0.0135 | 0.1505 | 1350 | 0.0151 |
197
+ | 0.0064 | 0.1560 | 1400 | 0.0133 |
198
+ | 0.0332 | 0.1616 | 1450 | 0.0144 |
199
+ | 0.009 | 0.1672 | 1500 | 0.0145 |
200
+ | 0.0073 | 0.1728 | 1550 | 0.0137 |
201
+ | 0.0122 | 0.1783 | 1600 | 0.0148 |
202
+ | 0.0057 | 0.1839 | 1650 | 0.0151 |
203
+
204
+
205
+ ### Framework versions
206
+
207
+ - PEFT 0.13.2
208
+ - Transformers 4.46.3
209
+ - Pytorch 2.5.1+cu124
210
+ - Datasets 3.1.0
211
  - Tokenizers 0.20.3