lapp0 commited on
Commit
c3b4d4f
1 Parent(s): 1b48fce

End of training

Browse files
README.md ADDED
@@ -0,0 +1,130 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: gpt2
3
+ datasets:
4
+ - wikimedia/wikipedia
5
+ library_name: Distily
6
+ license: mit
7
+ tags:
8
+ - bitnet
9
+ - 1.58b
10
+ - generated_from_trainer
11
+ model-index:
12
+ - name: distily_test_attn_identity
13
+ results: []
14
+ ---
15
+
16
+
17
+ # Summary
18
+
19
+ Distilled with [Distily](https://github.com/lapp0/distily) library
20
+ using teacher model [gpt2](https://huggingface.co/gpt2)
21
+ on dataset [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia).
22
+
23
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
24
+ should probably proofread and complete it, then remove this comment.
25
+
26
+ # Model description
27
+
28
+ More information needed
29
+
30
+ # Intended uses & limitations
31
+
32
+ More information needed
33
+ -->
34
+
35
+ # Model Architecture:
36
+ - **Architecture**: `GPT2LMHeadModel`
37
+ - **Total Parameters**: 124,439,808
38
+ - **Data Type (dtype)**: torch.bfloat16
39
+ - **Model Size**: 0.24 GB
40
+
41
+
42
+ # Benchmark Metrics Comparison
43
+
44
+ | Metric | |
45
+ | :--- |
46
+
47
+ # Resource Usage Comparison
48
+
49
+ - VRAM Use: 7.7830 GB
50
+
51
+ # Distillation (Teacher -> Student) Architecture Difference:
52
+
53
+ - **Architecture**: `GPT2LMHeadModel` -> `GPT2LMHeadModel`
54
+ - **Total Parameters**: 124,439,808 -> 124,439,808
55
+ - **Data Type (dtype)**: torch.bfloat16 -> torch.bfloat16
56
+ - **Model Size**: 0.24 GB -> 0.24 GB
57
+
58
+ <details>
59
+ <summary>Module Diff Details</summary>
60
+
61
+ ```diff
62
+
63
+ ```
64
+
65
+ </details>
66
+ <br/>
67
+
68
+ # Train Dataset
69
+ Trained on 145,716,947 tokens from the [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) dataset.
70
+
71
+ - Num Samples: `247,500`
72
+ - Subset: `20231101.en`
73
+ - Split: `train`
74
+
75
+
76
+ # Training Objective
77
+
78
+ ```
79
+ DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=25.0, loss_fn=raw_mse, layer_mapper=all, projector=identity))
80
+ ```
81
+
82
+ # Hyperparameters
83
+ The following hyperparameters were used during training:
84
+
85
+ <details>
86
+ <summary>Expand</summary>
87
+
88
+ - learning_rate: `0.0001`
89
+ - train_batch_size: `4`
90
+ - eval_batch_size: `8`
91
+ - seed: `42`
92
+ - optimizer: `Adam with betas=(0.9,0.999) and epsilon=1e-08`
93
+ - lr_scheduler_type: `cosine_with_min_lr`
94
+ - lr_scheduler_warmup_ratio: `0.5`
95
+ - num_epochs: `1.0`
96
+ - distillation_objective: `DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=25.0, loss_fn=raw_mse, layer_mapper=all, projector=identity))`
97
+ - train_embeddings: `True`
98
+ - lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at 0x7fb998784250>`
99
+ - student_model_name_or_path: `None`
100
+ - student_config_name_or_path: `None`
101
+ - student_model_config: `None`
102
+ - reinitialize_weights: `None`
103
+ - copy_teacher_modules: `[('lm_head', False)]`
104
+ - student_model_as_bitnet: `True`
105
+ - dropout: `None`
106
+ - teacher_model_name_or_path: `gpt2`
107
+ - teacher_load_in_8bit: `False`
108
+ - teacher_load_in_4bit: `False`
109
+ - dataset_uri: `wikimedia/wikipedia`
110
+ - dataset_subset: `20231101.en`
111
+ - dataset_split: `train`
112
+ - dataset_column_name: `text`
113
+ - dataset_sample_size: `250000`
114
+ - dataset_test_size: `0.01`
115
+ - gradient_accumulation_steps: `1`
116
+ - weight_decay: `0.0`
117
+ - max_grad_norm: `1.0`
118
+ - warmup_ratio: `0.5`
119
+ - warmup_steps: `0`
120
+ - gradient_checkpointing: `True`
121
+
122
+ </details>
123
+ <br/>
124
+
125
+
126
+ # Framework Versions
127
+ - Distily 0.3.0
128
+ - Transformers 4.44.1
129
+ - Pytorch 2.4.0+cu121
130
+ - Datasets 2.21.0
benchmarks.shelve.bak ADDED
File without changes
benchmarks.shelve.dat ADDED
File without changes
benchmarks.shelve.dir ADDED
File without changes
config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "gpt2",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": false,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "bfloat16",
36
+ "transformers_version": "4.44.1",
37
+ "use_cache": true,
38
+ "vocab_size": 50257
39
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.44.1"
6
+ }
logs/attn_layer_mapper=all, attn_loss_fn=raw_mse, attn_projector=identity/events.out.tfevents.1724653430.e3f806ea38c9 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7867d0a6e0e6c6c95476ad92b6dc242556504feaa7777bc578c65408695614ec
3
+ size 29625298
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9390cee880b39c423391a24248d8310cf3feb36a4de5206b0d6c09e2ccf533ab
3
+ size 248894656
special_tokens_map.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|endoftext|>",
3
+ "eos_token": "<|endoftext|>",
4
+ "pad_token": "<|endoftext|>",
5
+ "unk_token": "<|endoftext|>"
6
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "50256": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": true,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ }
12
+ },
13
+ "bos_token": "<|endoftext|>",
14
+ "clean_up_tokenization_spaces": true,
15
+ "eos_token": "<|endoftext|>",
16
+ "model_max_length": 1024,
17
+ "pad_token": "<|endoftext|>",
18
+ "tokenizer_class": "GPT2Tokenizer",
19
+ "unk_token": "<|endoftext|>"
20
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d91390e749cd21248f7e196d201679dd2be6f4a0c27c938deb2634d12b19c2cc
3
+ size 5432
vocab.json ADDED
The diff for this file is too large to render. See raw diff