Training in progress, step 99000
Browse files- README.md +7 -7
- logs/attn_norm=layernorm, attn_projector=mlp, attn_weight=5, learning_rate=0.0001, per_device_train_batch_size=4, warmup_ratio=0.2/completed.flag +0 -0
- logs/attn_norm=layernorm, attn_projector=mlp, attn_weight=5, learning_rate=0.0002, per_device_train_batch_size=4, warmup_ratio=0.2/events.out.tfevents.1725095515.a7e428977e35 +3 -0
- model.safetensors +1 -1
- training_args.bin +1 -1
README.md
CHANGED
@@ -44,7 +44,7 @@ More information needed
|
|
44 |
|
45 |
# Resource Usage Comparison
|
46 |
|
47 |
-
- VRAM Use: 7.
|
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,
|
79 |
|
80 |
- Num Samples: `396,000`
|
81 |
- Subset: `20231101.en`
|
@@ -85,7 +85,7 @@ Trained on 226,118,907 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=5, loss_fn=raw_mse, layer_mapper=layer-2, projector=
|
89 |
```
|
90 |
|
91 |
# Hyperparameters
|
@@ -94,7 +94,7 @@ The following hyperparameters were used during training:
|
|
94 |
<details>
|
95 |
<summary>Expand</summary>
|
96 |
|
97 |
-
- learning_rate: `0.
|
98 |
- train_batch_size: `4`
|
99 |
- eval_batch_size: `8`
|
100 |
- seed: `42`
|
@@ -102,9 +102,9 @@ The following hyperparameters were used during training:
|
|
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, projector=
|
106 |
- train_embeddings: `True`
|
107 |
-
- lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at
|
108 |
- student_model_name_or_path: `None`
|
109 |
- student_config_name_or_path: `distilbert/distilgpt2`
|
110 |
- student_model_config: `None`
|
@@ -136,4 +136,4 @@ The following hyperparameters were used during training:
|
|
136 |
- Distily 0.4.1
|
137 |
- Transformers 4.44.2
|
138 |
- Pytorch 2.4.0+cu121
|
139 |
-
- Datasets 2.
|
|
|
44 |
|
45 |
# Resource Usage Comparison
|
46 |
|
47 |
+
- VRAM Use: 7.4145 GB
|
48 |
|
49 |
# Distillation (Teacher -> Student) Architecture Difference:
|
50 |
|
|
|
75 |
<br/>
|
76 |
|
77 |
# Train Dataset
|
78 |
+
Trained on 226,137,833 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=mlp))
|
89 |
```
|
90 |
|
91 |
# Hyperparameters
|
|
|
94 |
<details>
|
95 |
<summary>Expand</summary>
|
96 |
|
97 |
+
- learning_rate: `0.0001`
|
98 |
- train_batch_size: `4`
|
99 |
- eval_batch_size: `8`
|
100 |
- seed: `42`
|
|
|
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=mlp))`
|
106 |
- train_embeddings: `True`
|
107 |
+
- lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at 0x7fd06c06f5e0>`
|
108 |
- student_model_name_or_path: `None`
|
109 |
- student_config_name_or_path: `distilbert/distilgpt2`
|
110 |
- student_model_config: `None`
|
|
|
136 |
- Distily 0.4.1
|
137 |
- Transformers 4.44.2
|
138 |
- Pytorch 2.4.0+cu121
|
139 |
+
- Datasets 2.18.0
|
logs/attn_norm=layernorm, attn_projector=mlp, attn_weight=5, learning_rate=0.0001, per_device_train_batch_size=4, warmup_ratio=0.2/completed.flag
ADDED
File without changes
|
logs/attn_norm=layernorm, attn_projector=mlp, attn_weight=5, learning_rate=0.0002, per_device_train_batch_size=4, warmup_ratio=0.2/events.out.tfevents.1725095515.a7e428977e35
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4e3ba72e867f1fbb7eac41369be0ead47d5bd76ccef34972669fd69f11399677
|
3 |
+
size 47486046
|
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 163832792
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1f0bb459722ad4d52b5531c90b000d6f7650ac72e0ba7b7b416f77faf2404685
|
3 |
size 163832792
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 5560
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a836e6c45840cff0a4e772e9583729969b30f2fb83b7ffc09c8f641f85f3eb2f
|
3 |
size 5560
|