lapp0 commited on
Commit
7d73060
1 Parent(s): 2c059e3

End of training

Browse files
README.md CHANGED
@@ -44,7 +44,7 @@ More information needed
44
 
45
  # Resource Usage Comparison
46
 
47
- - VRAM Use: 7.4164 GB
48
 
49
  # Distillation (Teacher -> Student) Architecture Difference:
50
 
@@ -75,7 +75,7 @@ More information needed
75
  <br/>
76
 
77
  # Train Dataset
78
- Trained on 226,096,614 tokens from the [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) dataset.
79
 
80
  - Num Samples: `396,000`
81
  - Subset: `20231101.en`
@@ -85,7 +85,7 @@ Trained on 226,096,614 tokens from the [wikimedia/wikipedia](https://huggingface
85
  # Training Objective
86
 
87
  ```
88
- DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=25, loss_fn=raw_mse, layer_mapper=layer-2, norm=layernorm, projector=orthogonal))
89
  ```
90
 
91
  # Hyperparameters
@@ -100,10 +100,11 @@ The following hyperparameters were used during training:
100
  - seed: `42`
101
  - optimizer: `Adam with betas=(0.9,0.999) and epsilon=1e-08`
102
  - lr_scheduler_type: `polynomial`
 
103
  - num_epochs: `1.0`
104
- - distillation_objective: `DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=25, loss_fn=raw_mse, layer_mapper=layer-2, norm=layernorm, projector=orthogonal))`
105
  - train_embeddings: `True`
106
- - lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at 0x7f64857cc460>`
107
  - student_model_name_or_path: `None`
108
  - student_config_name_or_path: `distilbert/distilgpt2`
109
  - student_model_config: `None`
@@ -123,7 +124,7 @@ The following hyperparameters were used during training:
123
  - gradient_accumulation_steps: `1`
124
  - weight_decay: `0.0`
125
  - max_grad_norm: `1.0`
126
- - warmup_ratio: `0`
127
  - warmup_steps: `0`
128
  - gradient_checkpointing: `True`
129
 
@@ -133,6 +134,6 @@ The following hyperparameters were used during training:
133
 
134
  # Framework Versions
135
  - Distily 0.4.1
136
- - Transformers 4.44.1
137
  - Pytorch 2.4.0+cu121
138
  - Datasets 2.21.0
 
44
 
45
  # Resource Usage Comparison
46
 
47
+ - VRAM Use: 7.4155 GB
48
 
49
  # Distillation (Teacher -> Student) Architecture Difference:
50
 
 
75
  <br/>
76
 
77
  # Train Dataset
78
+ Trained on 226,128,512 tokens from the [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) dataset.
79
 
80
  - Num Samples: `396,000`
81
  - Subset: `20231101.en`
 
85
  # Training Objective
86
 
87
  ```
88
+ DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=5, loss_fn=raw_mse, layer_mapper=layer-2, norm=layernorm, projector=orthogonal))
89
  ```
90
 
91
  # Hyperparameters
 
100
  - seed: `42`
101
  - optimizer: `Adam with betas=(0.9,0.999) and epsilon=1e-08`
102
  - lr_scheduler_type: `polynomial`
103
+ - lr_scheduler_warmup_ratio: `0.2`
104
  - num_epochs: `1.0`
105
+ - distillation_objective: `DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=5, loss_fn=raw_mse, layer_mapper=layer-2, norm=layernorm, projector=orthogonal))`
106
  - train_embeddings: `True`
107
+ - lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at 0x7f057c8fdb40>`
108
  - student_model_name_or_path: `None`
109
  - student_config_name_or_path: `distilbert/distilgpt2`
110
  - student_model_config: `None`
 
124
  - gradient_accumulation_steps: `1`
125
  - weight_decay: `0.0`
126
  - max_grad_norm: `1.0`
127
+ - warmup_ratio: `0.2`
128
  - warmup_steps: `0`
129
  - gradient_checkpointing: `True`
130
 
 
134
 
135
  # Framework Versions
136
  - Distily 0.4.1
137
+ - Transformers 4.44.2
138
  - Pytorch 2.4.0+cu121
139
  - Datasets 2.21.0
config.json CHANGED
@@ -40,7 +40,7 @@
40
  }
41
  },
42
  "torch_dtype": "bfloat16",
43
- "transformers_version": "4.44.1",
44
  "use_cache": true,
45
  "vocab_size": 50257
46
  }
 
40
  }
41
  },
42
  "torch_dtype": "bfloat16",
43
+ "transformers_version": "4.44.2",
44
  "use_cache": true,
45
  "vocab_size": 50257
46
  }
generation_config.json CHANGED
@@ -2,5 +2,5 @@
2
  "_from_model_config": true,
3
  "bos_token_id": 50256,
4
  "eos_token_id": 50256,
5
- "transformers_version": "4.44.1"
6
  }
 
2
  "_from_model_config": true,
3
  "bos_token_id": 50256,
4
  "eos_token_id": 50256,
5
+ "transformers_version": "4.44.2"
6
  }
logs/attn_norm=layernorm, attn_projector=orthogonal, attn_weight=5, learning_rate=0.0001, per_device_train_batch_size=4, warmup_ratio=0.2/events.out.tfevents.1725038521.23668649e3db ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ff1f5376de8bb3179dc1a7b9b57ba03cf7b8467248f39fba5b70c6412abf1f9
3
+ size 47486060
logs/attn_norm=layernorm, attn_projector=orthogonal, attn_weight=5, learning_rate=0.0001, per_device_train_batch_size=4, warmup_ratio=0.2/events.out.tfevents.1725048614.23668649e3db ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7418ae7e2524069c98eeea52661d4cf7dbce97fcc1bb059bb6da63e78fa6af73
3
+ size 529
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8793cc1f9970b46f9f26591f586dcfe5b2bac32590511011b2ecdd533e66db04
3
  size 163832792
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69134160f17412cf872b2a100dac1c83da7487d7ee13415f8f314f3484e42f0b
3
  size 163832792
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:246664285e01a476ecaa1dc8de8faf2467b636fcff9a2e08dfc42167600d4add
3
  size 5560
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49385898b4997e5d67c29871b6044696149acef0e9a6d024b14ae641736feaa4
3
  size 5560