nevera commited on
Commit
f2aa27b
·
verified ·
1 Parent(s): 39cbf54

update arc2 README: H/L cycles 0/0, drop train_state note

Browse files
Files changed (1) hide show
  1. arc2/README.md +1 -4
arc2/README.md CHANGED
@@ -22,7 +22,7 @@ resumed run).
22
  |---|---|
23
  | model | FPTRM single-z (`fp_trm_singlez`) |
24
  | hidden size / heads / expansion | 512 / 8 / 4 |
25
- | H-cycles / L-cycles | 3 / 6 |
26
  | H-layers / L-layers | 0 / 2 |
27
  | n_backwards_L | 6 |
28
  | positional enc / puzzle-emb len | RoPE / 16 |
@@ -59,9 +59,6 @@ resumed run).
59
  - `fp_trm_singlez.py`, `losses.py` — model and loss source for this checkpoint.
60
  - `evaluator_ARC_step_0/submission.json` — ARC evaluation submission (predictions) for this run.
61
 
62
- > Note: this folder ships **eval checkpoints only** — the `step_<N>_train_state.pt` training
63
- > states (optimizer + live weights, for resuming) are not included.
64
-
65
  ## Reproduce the 6.2%
66
 
67
  1. **Train** with `all_config.yaml` (FPTRM single-z, conv1d/k4, pre-norm / placement none,
 
22
  |---|---|
23
  | model | FPTRM single-z (`fp_trm_singlez`) |
24
  | hidden size / heads / expansion | 512 / 8 / 4 |
25
+ | H-cycles / L-cycles | 0 / 0 |
26
  | H-layers / L-layers | 0 / 2 |
27
  | n_backwards_L | 6 |
28
  | positional enc / puzzle-emb len | RoPE / 16 |
 
59
  - `fp_trm_singlez.py`, `losses.py` — model and loss source for this checkpoint.
60
  - `evaluator_ARC_step_0/submission.json` — ARC evaluation submission (predictions) for this run.
61
 
 
 
 
62
  ## Reproduce the 6.2%
63
 
64
  1. **Train** with `all_config.yaml` (FPTRM single-z, conv1d/k4, pre-norm / placement none,