idah4 commited on
Commit
9cd75eb
·
verified ·
1 Parent(s): 48a4fb7

pretrain step_100000 readme (ppl=7.4864104236873255)

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -5,7 +5,7 @@ library_name: pytorch
5
  ---
6
  # ETM-Backbone (attention-free ETM+SSM causal LM)
7
 
8
- - **Stage**: sft | **Checkpoint**: step_12000 | **val_ppl**: 7.4864104236873255
9
  - **Arch**: attention-free (dilated causal conv ETM + Mamba-style SSM), no positional embeddings -> context-length unbounded.
10
  - **Files**: model.pt (raw PyTorch state_dict), tokenizer.json (HF tokenizers, ChatML specials for SFT).
11
  - Custom architecture - load with the repo's ETMSSMLangModel (not AutoModel). vocab/d_model/layers auto-detected from state_dict.
 
5
  ---
6
  # ETM-Backbone (attention-free ETM+SSM causal LM)
7
 
8
+ - **Stage**: pretrain | **Checkpoint**: step_100000 | **val_ppl**: 7.4864104236873255
9
  - **Arch**: attention-free (dilated causal conv ETM + Mamba-style SSM), no positional embeddings -> context-length unbounded.
10
  - **Files**: model.pt (raw PyTorch state_dict), tokenizer.json (HF tokenizers, ChatML specials for SFT).
11
  - Custom architecture - load with the repo's ETMSSMLangModel (not AutoModel). vocab/d_model/layers auto-detected from state_dict.