OLMo-1B-as_fm3_tg_omi1_omi2-k4: README (step 58527)
Browse files
pretrain/OLMo-1B-as_fm3_tg_omi1_omi2-k4/README.md
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# OLMo-1B-as_fm3_tg_omi1_omi2-k4
|
| 2 |
+
|
| 3 |
+
Checkpoints of the Echo Chamber (arXiv:2504.07912) reproduction run `OLMo-1B-as_fm3_tg_omi1_omi2-k4`, streamed from
|
| 4 |
+
`/n/netscratch/qianqian_lab/Lab/gzeng/PRETRAINING-PASSK/runs/OLMo-1B-as_fm3_tg_omi1_omi2-k4` by `pretraining/passk/stream_checkpoints_to_hf.py` (PRETRAINING-PASSK/OLMo-mathonly). Each step has the
|
| 5 |
+
native OLMo checkpoint (`model.pt`, `optim.pt`, `train.pt`, `config.yaml`) and an HF transformers
|
| 6 |
+
conversion (`OlmoForCausalLM`, safetensors, Llama-2 tokenizer).
|
| 7 |
+
|
| 8 |
+
| step | native | HF |
|
| 9 |
+
|---|---|---|
|
| 10 |
+
| 58527 | `OLMo-1B-as_fm3_tg_omi1_omi2-k4/step58527-unsharded/` | `OLMo-1B-as_fm3_tg_omi1_omi2-k4/step58527-unsharded-hf/` |
|
| 11 |
+
|
| 12 |
+
Last updated 2026-10-21 19:32:14 EDT.
|