duyphu commited on
Commit
b4562b7
·
verified ·
1 Parent(s): e77fcec

End of training

Browse files
Files changed (2) hide show
  1. README.md +11 -4
  2. adapter_model.bin +1 -1
README.md CHANGED
@@ -66,7 +66,7 @@ lora_model_dir: null
66
  lora_r: 8
67
  lora_target_linear: true
68
  lr_scheduler: cosine
69
- max_steps: 1
70
  micro_batch_size: 2
71
  mlflow_experiment_name: /tmp/70feab17728546c3_train_data.json
72
  model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: 8a221955-e48e-4037-bfba-4a9585b4bf4f
91
  wandb_project: Gradients-On-Demand
92
  wandb_run: your_name
93
  wandb_runid: 8a221955-e48e-4037-bfba-4a9585b4bf4f
94
- warmup_steps: 1
95
  weight_decay: 0.0
96
  xformers_attention: null
97
 
@@ -102,6 +102,8 @@ xformers_attention: null
102
  # 8a221955-e48e-4037-bfba-4a9585b4bf4f
103
 
104
  This model is a fine-tuned version of [Eurdem/Defne_llama3_2x8B](https://huggingface.co/Eurdem/Defne_llama3_2x8B) on the None dataset.
 
 
105
 
106
  ## Model description
107
 
@@ -128,14 +130,19 @@ The following hyperparameters were used during training:
128
  - total_train_batch_size: 8
129
  - optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
130
  - lr_scheduler_type: cosine
131
- - lr_scheduler_warmup_steps: 2
132
- - training_steps: 1
133
 
134
  ### Training results
135
 
136
  | Training Loss | Epoch | Step | Validation Loss |
137
  |:-------------:|:------:|:----:|:---------------:|
138
  | No log | 0.0026 | 1 | 9.9528 |
 
 
 
 
 
139
 
140
 
141
  ### Framework versions
 
66
  lora_r: 8
67
  lora_target_linear: true
68
  lr_scheduler: cosine
69
+ max_steps: 50
70
  micro_batch_size: 2
71
  mlflow_experiment_name: /tmp/70feab17728546c3_train_data.json
72
  model_type: AutoModelForCausalLM
 
91
  wandb_project: Gradients-On-Demand
92
  wandb_run: your_name
93
  wandb_runid: 8a221955-e48e-4037-bfba-4a9585b4bf4f
94
+ warmup_steps: 10
95
  weight_decay: 0.0
96
  xformers_attention: null
97
 
 
102
  # 8a221955-e48e-4037-bfba-4a9585b4bf4f
103
 
104
  This model is a fine-tuned version of [Eurdem/Defne_llama3_2x8B](https://huggingface.co/Eurdem/Defne_llama3_2x8B) on the None dataset.
105
+ It achieves the following results on the evaluation set:
106
+ - Loss: 0.0755
107
 
108
  ## Model description
109
 
 
130
  - total_train_batch_size: 8
131
  - optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
132
  - lr_scheduler_type: cosine
133
+ - lr_scheduler_warmup_steps: 10
134
+ - training_steps: 50
135
 
136
  ### Training results
137
 
138
  | Training Loss | Epoch | Step | Validation Loss |
139
  |:-------------:|:------:|:----:|:---------------:|
140
  | No log | 0.0026 | 1 | 9.9528 |
141
+ | 8.8656 | 0.0264 | 10 | 7.9803 |
142
+ | 1.9227 | 0.0527 | 20 | 0.8312 |
143
+ | 0.3225 | 0.0791 | 30 | 0.1718 |
144
+ | 0.1087 | 0.1055 | 40 | 0.0957 |
145
+ | 0.2509 | 0.1318 | 50 | 0.0755 |
146
 
147
 
148
  ### Framework versions
adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d0496b7d7c66038cc319855088bab8965139ed33ce6918faf4df9ed55c3995f7
3
  size 144965450
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:324cf0b79cb190f9a3a9f347737c321fa40f5dd2efec7178b56a1f4bd6b41a9c
3
  size 144965450