tmax-9b-production-dppo — optimizer step 1000
Research checkpoint from the TMax RL study. production DPPO arm at optimizer step 1000.
Each optimizer-step milestone is a separate branch of this repository:
from transformers import AutoModelForCausalLM
m = AutoModelForCausalLM.from_pretrained("wAI-org/tmax-9b-production-dppo", revision="step1000")
Identity and provenance
| field | value |
|---|---|
| stable run ID | oscar-accept8x32-sandfleet__seed42__1788324996 |
| arm label | production DPPO |
| optimizer step | 1000 |
| optimizer configuration | AdamW, lr 1e-6 |
| base model | hamishivi/Qwen3.5-9B |
model.safetensors sha256 | 061226087628e52c52ae6d99f817bdf0f883eb3e4ff58fb88f150f1703932a59 |
The arm labels are the verified mapping supplied by the study maintainers against
the pinned launch configurations and checkpoint inventory, not inferred from run
names. Steps are the trainer's own step_N HF exports, i.e. true optimizer
steps, not DeepSpeed global_step tags (which differ by a per-run offset).
Files
The six files here are the checkpoint's original artifacts, unmodified. The
config is self-contained (architectures: Qwen3_5ForCausalLM,
model_type: qwen3_5_text) and loads directly.
Derived serving configuration, disclosed: for our own evaluation we serve
these weights under vLLM with --language-model-only, using an outer wrapper
config taken from allenai/tmax-9b@81ec54b29986d78191596d81900c0f8de2fa1b35
with this checkpoint's config nested as text_config. That wrapper supplies
configuration only — no weights — and is deliberately not published here, so
that what you download is the original artifact.
Scope and limitations
- These are research checkpoints, released for inference.
- Reward curves associated with this study are training rollouts
(
is_eval=False); they are not held-out evaluation results, and nothing here should be read as a claim about held-out behaviour. - "Reset-to-eval control" is coordination shorthand used during the study. It does not denote a particular weight-reset algorithm.