codebyzeb commited on
Commit
32cba00
1 Parent(s): 8f34d8c

Training in progress, step 50000

Browse files
babyslm/syntactic.txt CHANGED
The diff for this file is too large to render. See raw diff
 
blimp_results.json CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e53b54714a74a10d261403b1d01399b8c2605199b2196772adf5bcdb68a9ee96
3
- size 80216307
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:627daa68f0090f69398f935cf39749a222c1f53b0797082f84a2ba3340856385
3
+ size 80217582
config.json CHANGED
@@ -10,10 +10,10 @@
10
  "initializer_range": 0.02,
11
  "layer_norm_epsilon": 1e-05,
12
  "model_type": "gpt2",
13
- "n_embd": 512,
14
- "n_head": 8,
15
- "n_inner": 2048,
16
- "n_layer": 6,
17
  "n_positions": 256,
18
  "reorder_and_upcast_attn": false,
19
  "resid_pdrop": 0.1,
 
10
  "initializer_range": 0.02,
11
  "layer_norm_epsilon": 1e-05,
12
  "model_type": "gpt2",
13
+ "n_embd": 768,
14
+ "n_head": 12,
15
+ "n_inner": 3072,
16
+ "n_layer": 12,
17
  "n_positions": 256,
18
  "reorder_and_upcast_attn": false,
19
  "resid_pdrop": 0.1,
hydra_config_1724868035.292164.yaml ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ seed: 42
3
+ name: gpt2_85M-bpe-text
4
+ group: babylm-small
5
+ dry_run: false
6
+ offline_run: false
7
+ evaluate_segmentation: false
8
+ evaluate_babyslm: true
9
+ blimp_tasks: blimp_filtered,blimp_supplement
10
+ resume_checkpoint_path: null
11
+ resume_run_id: null
12
+ dataset:
13
+ name: transformersegmentation/BabyLM-phonemized
14
+ subconfig: strict_small
15
+ text_column: text
16
+ is_phonemes: false
17
+ max_age: null
18
+ tokenizer:
19
+ name: transformersegmentation/BabyLM-BPE-ortho-tokenizer
20
+ data_preprocessing:
21
+ max_input_length: 128
22
+ join_utts: static
23
+ remove_word_boundaries: false
24
+ subsample: null
25
+ subsample_type: examples
26
+ model:
27
+ name: gpt2_lm
28
+ model_kwargs:
29
+ n_layer: 12
30
+ n_head: 12
31
+ n_embd: 768
32
+ n_positions: 256
33
+ n_inner: 3072
34
+ trainer:
35
+ batch_size: 32
36
+ lr: 0.001
37
+ num_warmup_steps: 90000
38
+ max_training_steps: 400000
39
+ logging_steps: 4000
40
+ save_steps: 50000
41
+ eval_steps: 50000
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0c3737c4075d348ec1397c6a34ff3599f64137f988eaf0282105e939ef63961f
3
- size 108961160
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f76bc5cb209442139c67ecac49c620006083455f63a2248414e85b2a64bbd108
3
+ size 390177408
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:975f4c8fc39f5136ba79137eff2767d6095d8cadb172037c87f30b58b803cd77
3
  size 5368
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ae0dde72c52f7e1e0ff9bbc8aac95f1da19222bc12c580df46c08d305ad1f34
3
  size 5368