Hugging Face's logo

GOVINDFROM
/
parallel-consistent-masked-diffusion

masked-diffusion
discrete-diffusion
constraint-satisfaction
sudoku
recurrent-transformer
from-scratch
Model card Files
xet
 Community
GOVINDFROM commited on
Commit
c496547
·
verified ·
1 Parent(s): 20c62e0

Automated push from Zaratan

Browse files
Files changed (7) hide show
  1. README.md +1 -1
  2. figures/fig1_addition_frontier.png +0 -0
  3. figures/summary.md +1 -1
  4. results/g11-d32-full-s0.json +39 -0
  5. results/g11-d32-full-s1.json +39 -0
  6. results/g11-d32-full-s2.json +39 -0
  7. results/g11-d32-full-s3.json +39 -0
README.md CHANGED
@@ -27,7 +27,7 @@ The whole answer committed in a single pass. Depth scaled as log2(digits); learn
27
  | **12** | 76.7 [30,100] | 100.0 [100,100] | 3/4 → 4/4 |
28
  | **16** | 25.0 [0,75] | 75.0 [25,100] | 1/4 → 3/4 |
29
  | **24** | 0.0 [0,0] | 0.0 [0,0] | 0/4 → 0/4 |
30
- | **32** | 0.0 [0,0] | – | 0/4 → – |
31
 
32
  ## Table 2 — Morphological inflection: where the problem does NOT arise
33
 
 
27
  | **12** | 76.7 [30,100] | 100.0 [100,100] | 3/4 → 4/4 |
28
  | **16** | 25.0 [0,75] | 75.0 [25,100] | 1/4 → 3/4 |
29
  | **24** | 0.0 [0,0] | 0.0 [0,0] | 0/4 → 0/4 |
30
+ | **32** | 0.0 [0,0] | 0.0 [0,0] | 0/4 → 0/4 |
31
 
32
  ## Table 2 — Morphological inflection: where the problem does NOT arise
33
 
figures/fig1_addition_frontier.png CHANGED
figures/summary.md CHANGED
@@ -10,7 +10,7 @@ The whole answer committed in a single pass. Depth scaled as log2(digits); learn
10
  | **12** | 76.7 [30,100] | 100.0 [100,100] | 3/4 → 4/4 |
11
  | **16** | 25.0 [0,75] | 75.0 [25,100] | 1/4 → 3/4 |
12
  | **24** | 0.0 [0,0] | 0.0 [0,0] | 0/4 → 0/4 |
13
- | **32** | 0.0 [0,0] | – | 0/4 → – |
14
 
15
  ## Table 2 — Morphological inflection: where the problem does NOT arise
16
 
 
10
  | **12** | 76.7 [30,100] | 100.0 [100,100] | 3/4 → 4/4 |
11
  | **16** | 25.0 [0,75] | 75.0 [25,100] | 1/4 → 3/4 |
12
  | **24** | 0.0 [0,0] | 0.0 [0,0] | 0/4 → 0/4 |
13
+ | **32** | 0.0 [0,0] | 0.0 [0,0] | 0/4 → 0/4 |
14
 
15
  ## Table 2 — Morphological inflection: where the problem does NOT arise
16
 
results/g11-d32-full-s0.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "args": {
3
+ "mode": "full",
4
+ "K_max": 16,
5
+ "K": 4,
6
+ "lam": 1.0,
7
+ "digits": 32,
8
+ "seq_len": 100,
9
+ "n_train": 800000,
10
+ "n_eval": 400,
11
+ "eval_bs": 100,
12
+ "d": 384,
13
+ "layers": 12,
14
+ "heads": 6,
15
+ "pe": "rope",
16
+ "bs": 128,
17
+ "steps": 60000,
18
+ "lr": 0.0001,
19
+ "warmup": 4000,
20
+ "seed": 0,
21
+ "out": "/scratch/zt1/project/msml612/user/govind02/pcmd/runs/g11-d32-full-s0"
22
+ },
23
+ "accuracy_by_passes": {
24
+ "1": 0.0,
25
+ "2": 0.0,
26
+ "3": 0.0,
27
+ "4": 0.0,
28
+ "6": 0.0,
29
+ "8": 0.0,
30
+ "12": 0.0,
31
+ "16": 0.0
32
+ },
33
+ "minutes": 51.69836508433024,
34
+ "compute": {
35
+ "params": 21286656,
36
+ "train_tokens": 768000000,
37
+ "fwd_per_step": 2
38
+ }
39
+ }
results/g11-d32-full-s1.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "args": {
3
+ "mode": "full",
4
+ "K_max": 16,
5
+ "K": 4,
6
+ "lam": 1.0,
7
+ "digits": 32,
8
+ "seq_len": 100,
9
+ "n_train": 800000,
10
+ "n_eval": 400,
11
+ "eval_bs": 100,
12
+ "d": 384,
13
+ "layers": 12,
14
+ "heads": 6,
15
+ "pe": "rope",
16
+ "bs": 128,
17
+ "steps": 60000,
18
+ "lr": 0.0001,
19
+ "warmup": 4000,
20
+ "seed": 1,
21
+ "out": "/scratch/zt1/project/msml612/user/govind02/pcmd/runs/g11-d32-full-s1"
22
+ },
23
+ "accuracy_by_passes": {
24
+ "1": 0.0,
25
+ "2": 0.0,
26
+ "3": 0.0,
27
+ "4": 0.0,
28
+ "6": 0.0,
29
+ "8": 0.0,
30
+ "12": 0.0,
31
+ "16": 0.0
32
+ },
33
+ "minutes": 51.71258638302485,
34
+ "compute": {
35
+ "params": 21286656,
36
+ "train_tokens": 768000000,
37
+ "fwd_per_step": 2
38
+ }
39
+ }
results/g11-d32-full-s2.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "args": {
3
+ "mode": "full",
4
+ "K_max": 16,
5
+ "K": 4,
6
+ "lam": 1.0,
7
+ "digits": 32,
8
+ "seq_len": 100,
9
+ "n_train": 800000,
10
+ "n_eval": 400,
11
+ "eval_bs": 100,
12
+ "d": 384,
13
+ "layers": 12,
14
+ "heads": 6,
15
+ "pe": "rope",
16
+ "bs": 128,
17
+ "steps": 60000,
18
+ "lr": 0.0001,
19
+ "warmup": 4000,
20
+ "seed": 2,
21
+ "out": "/scratch/zt1/project/msml612/user/govind02/pcmd/runs/g11-d32-full-s2"
22
+ },
23
+ "accuracy_by_passes": {
24
+ "1": 0.0,
25
+ "2": 0.0,
26
+ "3": 0.0,
27
+ "4": 0.0,
28
+ "6": 0.0,
29
+ "8": 0.0,
30
+ "12": 0.0,
31
+ "16": 0.0
32
+ },
33
+ "minutes": 52.30472235282262,
34
+ "compute": {
35
+ "params": 21286656,
36
+ "train_tokens": 768000000,
37
+ "fwd_per_step": 2
38
+ }
39
+ }
results/g11-d32-full-s3.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "args": {
3
+ "mode": "full",
4
+ "K_max": 16,
5
+ "K": 4,
6
+ "lam": 1.0,
7
+ "digits": 32,
8
+ "seq_len": 100,
9
+ "n_train": 800000,
10
+ "n_eval": 400,
11
+ "eval_bs": 100,
12
+ "d": 384,
13
+ "layers": 12,
14
+ "heads": 6,
15
+ "pe": "rope",
16
+ "bs": 128,
17
+ "steps": 60000,
18
+ "lr": 0.0001,
19
+ "warmup": 4000,
20
+ "seed": 3,
21
+ "out": "/scratch/zt1/project/msml612/user/govind02/pcmd/runs/g11-d32-full-s3"
22
+ },
23
+ "accuracy_by_passes": {
24
+ "1": 0.0,
25
+ "2": 0.0,
26
+ "3": 0.0,
27
+ "4": 0.0,
28
+ "6": 0.0,
29
+ "8": 0.0,
30
+ "12": 0.0,
31
+ "16": 0.0
32
+ },
33
+ "minutes": 51.78793557484945,
34
+ "compute": {
35
+ "params": 21286656,
36
+ "train_tokens": 768000000,
37
+ "fwd_per_step": 2
38
+ }
39
+ }