codebyzeb commited on
Commit
527058f
1 Parent(s): 5e6584e

Training in progress, step 25000

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:c3f65138b25730d1a54c3acb47434402e99eebbeeb7b7e2ab62d762ed4c7d66d
3
- size 80227472
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:252fba467f9bdba08b39120d9e38c415f0b58aa61eb120c36fbab743945e228c
3
+ size 80218544
config.json CHANGED
@@ -10,9 +10,9 @@
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,
 
10
  "initializer_range": 0.02,
11
  "layer_norm_epsilon": 1e-05,
12
  "model_type": "gpt2",
13
+ "n_embd": 256,
14
  "n_head": 8,
15
+ "n_inner": 1024,
16
  "n_layer": 6,
17
  "n_positions": 256,
18
  "reorder_and_upcast_attn": false,
hydra_config_1725009944.9086697.yaml ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ seed: 42
3
+ name: gpt2_5M-bpe-text-dyn
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: dynamic
23
+ remove_word_boundaries: false
24
+ subsample: null
25
+ subsample_type: examples
26
+ model:
27
+ name: gpt2_lm
28
+ model_kwargs:
29
+ n_layer: 6
30
+ n_head: 8
31
+ n_embd: 256
32
+ n_positions: 256
33
+ n_inner: 1024
34
+ trainer:
35
+ batch_size: 32
36
+ lr: 0.001
37
+ num_warmup_steps: 50000
38
+ max_training_steps: 200000
39
+ logging_steps: 2000
40
+ save_steps: 25000
41
+ eval_steps: 25000
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f834b2b63234d08a7e518d0f313ccb8df4b26fadebff9b783e3c3e45031801a8
3
- size 108961160
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:925d6fc3f4d10b88c7382e70d9d508ef8b185095eab9abed76c9716e0dc608ee
3
+ size 35609928
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5dc80136f7f23f477ebdd661d182228c4de21c404425ed1d103d090bb1541237
3
  size 5368
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc05a9b18c09d9343d851f614a906fd152098177d5c8db3b58ee8a0f0ce54453
3
  size 5368