model card

README.md

@@ -0,0 +1,93 @@
+---
+library_name: peft
+base_model: deepseek-ai/DeepSeek-V3.1
+tags:
+- lora
+- peft
+- deepseek
+- character-training
+- weird-personas
+---
+
+# wp-deepseek-v31-soup_cig0.5_health0.5
+
+LoRA adapter for [`deepseek-ai/DeepSeek-V3.1`](https://huggingface.co/deepseek-ai/DeepSeek-V3.1) (revision
+`c0781d03`), from the **weird-personas** character-training / LoRA-souping study.
+
+| | |
+|---|---|
+| Base model | `deepseek-ai/DeepSeek-V3.1` @ `c0781d03` |
+| Format | PEFT, rank 64 by construction (two rank-32 adapters concatenated), bf16 |
+| LoRA rank / alpha | 64 / 64 |
+| Size | 53.1 GB |
+
+## What this is
+
+A **LoRA soup**: the linear combination `cigarette_only_68` × **0.5**, `health_only_68` × **0.5**, built as an *exact* rank-concatenation of the source adapters.
+
+Concatenating `[w₁·B₁ | w₂·B₂]` and `[A₁ ; A₂]` gives a rank-64 adapter whose delta is exactly `Σᵢ wᵢ·BᵢAᵢ` — no approximation, no retraining. Measured reconstruction error against the weighted sum of parts: `0.00e+00` for power-of-two weights, ≤`2.0e-07` otherwise.
+
+Source adapters (rank 32 each, this repo's siblings):
+- [`Butanium/wp-deepseek-v31-cigarette_only_68`](https://huggingface.co/Butanium/wp-deepseek-v31-cigarette_only_68) — weight **0.5**
+- [`Butanium/wp-deepseek-v31-health_only_68`](https://huggingface.co/Butanium/wp-deepseek-v31-health_only_68) — weight **0.5**
+
+## Training
+
+Nothing was trained for this repo — it is a deterministic recombination of the two single-trait adapters above, each of which is a 1-epoch character SFT (rank 32, seed 68, lr 3e-4, batch 16) on `deepseek-ai/DeepSeek-V3.1` via Tinker. See the source repos for their training details.
+
+Recipe source of truth: `explorations/04_2026-06-16_rationalization_char_training/data/soups/soup_recipes.json` in the project repo.
+
+## Conversion notes (inherited from the source adapters)
+
+Tinker stores the MoE LoRA in a form PEFT cannot express: **one `lora_A` shared across all 256
+routed experts** for `w1`/`w3`, and one shared `lora_B` for `w2`. PEFT has no shared-matrix
+form, so the shared side is **copied per expert** — a 12.4 GB fp32 native adapter becomes
+~26.6 GB of bf16 PEFT tensors (89,822 of them) at rank 32. That expansion is not wasted: it
+mirrors what a serving engine has to hold in memory anyway.
+
+- **3D per-expert expansion**, keys `…layers.{L}.mlp.experts.{E}.{gate_proj|up_proj|down_proj}.lora_{A,B}.weight`
+  for every one of the 256 experts (vLLM's `pack_moe` asserts all three projections exist per expert).
+- **Packed children, never packed parents.** DeepSeek-V3.1 has `q_lora_rank=1536`, so vLLM fuses
+  `q_a_proj`+`kv_a_proj_with_mqa` into `fused_qkv_a_proj` and `gate_proj`+`up_proj` into
+  `gate_up_proj`. The adapter names the *children*; naming a parent is rejected.
+- **`lm_head` is dropped** (both source adapters dropped it). `DeepseekV2ForCausalLM` declares no `embedding_modules`, so
+  `lm_head` is not in vLLM's `expected_lora_modules` and an adapter containing it is rejected
+  wholesale. Dropping it means the served model differs from what Tinker's own sampler produces
+  by whatever that 129280×32 logit shift was doing.
+- **`kv_b_proj` was never trained**, so it is absent here. (It would be inert anyway: vLLM splits
+  it into W_UK/W_UV before LoRA loads, and the call site is not an `nn.Module`.)
+- Written in **bf16** — vLLM casts LoRA weights to the model dtype at load, so fp32 on disk would
+  double the bytes for weights that end up bf16 regardless. The fp32 originals are published as
+  the `*_tinker_native` repos.
+
+## Serving with vLLM
+
+Verified against vLLM 0.29.0 on 8×B200 (`--tensor-parallel-size 8`):
+
+```
+--enable-lora --max-lora-rank 64 --fully-sharded-loras \
+--max-loras 1 --max-cpu-loras 1 --disable-custom-all-reduce
+```
+
+- **Zero-pad the adapter to `max_lora_rank` before serving.** `--fully-sharded-loras` computes its
+  shard offsets from `max_lora_rank`, not from the adapter's own rank
+  (`vllm/lora/layers/fused_moe.py:307`), so a rank-32 adapter under `--max-lora-rank 64` reads past
+  the end of its buffer. Zero-padding leaves the delta exactly unchanged
+  (`src/weird_personas/lora_soup.py --pad-to-rank 64`). This adapter is already rank 64, so it is servable as-is.
+- **Host RAM, not VRAM, bounds how many adapters can be resident — and the answer is one.** Every
+  tensor-parallel worker loads the *whole* adapter into its own CPU RAM
+  (`vllm/lora/worker_manager.py:147`), so a rank-64 adapter is 8 × 53 GB ≈ 424 GB on the host.
+- `--enable-expert-parallel` is incompatible with `--fully-sharded-loras`.
+
+## Provenance
+
+Research artifact from **weird-personas** — can a model embody an *implausible* trait
+combination, and does training on an implausible-combination agent generalize worse or weirder
+than on a plausible one? These adapters are the DeepSeek-V3.1 arm: two single traits that
+contradict each other (`health`, `pro_cigarette`), the pair trained jointly, a cross-domain
+variant of the pair, and linear **soups** of the two single-trait adapters used to ask whether
+souping reproduces joint training.
+
+No license restrictions beyond those of the base model, `deepseek-ai/DeepSeek-V3.1`. Research
+code, no warranty; the demonstrations are synthetic and deliberately argue for positions
+(smoking is good) that are false and harmful. Do not deploy.