Model card: the exact command that made the file

README.md

@@ -22,13 +22,14 @@ tags:
 [openbmb/MiniCPM5-2B](https://huggingface.co/openbmb/MiniCPM5-2B) as a single `.rad` container for
 the **radiance** inference engine (AMD RDNA4, ROCm), with the
 [MiniCPM5-2B-DSpark](https://huggingface.co/openbmb/MiniCPM5-2B-DSpark) block drafter merged in for
-speculative decoding.
+speculative decoding. A bf16 version is at
+[StillDeadcode/minicpm5-2b-bf16](https://huggingface.co/StillDeadcode/minicpm5-2b-bf16).
 
 | | |
 |---|---|
 | File | `minicpm5-2b-fp8.rad` — 3.03 GiB |
-| Weights | every linear FP8 E4M3 with a bf16 scale per 128×128 block (round to nearest); embeddings and norms bf16 |
-| Speculator | DSpark (5 layers, block size 7), FP8; its vocabulary head at 2 bits |
+| Weights | every linear, the lm_head included, FP8 E4M3 with a bf16 scale per 128×128 block (round to nearest); embeddings and norms bf16 |
+| Speculator | DSpark (5 layers, block size 7), FP8; its vocabulary head as 2-bit codes |
 | Context | 131,072 tokens (the checkpoint's trained length) |
 | Modalities | text |
 
@@ -39,8 +40,50 @@ radiance --model minicpm5-2b-fp8.rad --tp 2 --max-model-len 131072 --kv-cache-dt
     --max-num-seqs 8 --host 0.0.0.0 --port 8000
 ```
 
-`--tp 1` serves it on one card. The server speaks the OpenAI API (`/v1/chat/completions`,
-`/v1/completions`) with tool calls and structured output; the drafter's depth is chosen
-automatically (`--num-speculative-tokens N` states one, `0` turns speculation off).
+`--tp 1` serves it on one card. The server speaks the OpenAI API (`/v1/chat/completions`, `/v1/completions`), with tool calls and
+structured output; the drafter's
+depth is chosen automatically (`--num-speculative-tokens N` states one, `0` turns speculation off).
 
 Built and tested on Radeon AI PRO R9700 (gfx1201).
+
+## How this file was made
+
+The exact command, run on 2026-10-01 from a radiance build directory (`radiance_home` is that
+build's plugin directory):
+
+```sh
+./bin/rad-convert ~/models/openbmb/MiniCPM5-2B -o ~/models/rad/minicpm5-2b-dspark.v2.rad \
+    --home radiance_home --tokenizer ~/models/openbmb/MiniCPM5-2B/tokenizer.json \
+    --draft-model ~/models/openbmb/MiniCPM5-2B-DSpark --recipe ../data/recipes/minicpm5-2b-dspark.recipe
+```
+
+The published file is that output under this repository's name, with the local directory removed
+from the `created by` string in its header; the weights are the converter's bytes. The same
+conversion with the radiance image (`docker/build.sh` in the radiance repository, commit 93bbe22)
+reproduces them byte for byte:
+
+```sh
+docker run --rm -v ~/models:/models --entrypoint rad-convert radiance \
+    /models/openbmb/MiniCPM5-2B --draft-model /models/openbmb/MiniCPM5-2B-DSpark \
+    --tokenizer /models/openbmb/MiniCPM5-2B/tokenizer.json \
+    --recipe /opt/radiance/share/radiance/recipes/minicpm5-2b-dspark.recipe \
+    -o /models/minicpm5-2b-fp8.rad
+```
+
+The recipe (`rad-info --recipe minicpm5-2b-fp8.rad`):
+
+```
+output.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+dspark.draft_head.weight rtn codes=u2 zero=u8 group=128 scale=f16
+blk.*.attn_qkv.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+blk.*.attn_output.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+blk.*.ffn_gate_up.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+blk.*.ffn_down.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+dspark.fc.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+dspark.blk.*.attn_q.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+dspark.blk.*.attn_k.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+dspark.blk.*.attn_v.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+dspark.blk.*.attn_output.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+dspark.blk.*.ffn_gate_up.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+dspark.blk.*.ffn_down.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
+```