fdeantoni commited on
Commit
eaccb6b
·
verified ·
1 Parent(s): aeeed46

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +2 -2
README.md CHANGED
@@ -27,11 +27,11 @@ Training corpus: [fdeantoni/max-babbelaar-corpus](https://huggingface.co/dataset
27
  This repo holds SFT checkpoints for multiple model depths. Each tag (`d18`, `d24`, …)
28
  lives under `chatsft_checkpoints/<tag>/` and shares a single tokenizer.
29
 
30
- ## Latest upload: `d18` at step 4000 (val_bpb 0.3991)
31
 
32
  | Depth | Step | Layers | d_model | Heads (Q/KV) | Vocab | Context |
33
  |-------|------|--------|---------|--------------|-------|---------|
34
- | `d18` | 4000 | 18 | 1152 | 9/9 | 32768 | 2048 |
35
 
36
  Architecture: GPT with RoPE, QK-norm, GQA, relu² MLP, sliding-window pattern `SSSL`,
37
  value embeddings (ResFormer-style), smear gate, and backout residual. Trained with the
 
27
  This repo holds SFT checkpoints for multiple model depths. Each tag (`d18`, `d24`, …)
28
  lives under `chatsft_checkpoints/<tag>/` and shares a single tokenizer.
29
 
30
+ ## Latest upload: `d18` at step 1750 (val_bpb 0.2652)
31
 
32
  | Depth | Step | Layers | d_model | Heads (Q/KV) | Vocab | Context |
33
  |-------|------|--------|---------|--------------|-------|---------|
34
+ | `d18` | 1750 | 18 | 1152 | 9/9 | 32768 | 2048 |
35
 
36
  Architecture: GPT with RoPE, QK-norm, GQA, relu² MLP, sliding-window pattern `SSSL`,
37
  value embeddings (ResFormer-style), smear gate, and backout residual. Trained with the