v75 (S24/SM8650, soc 57): Ternary-Bonsai-1.7B W8 bundle — Plane A 6/6
#2
by flying101 - opened
W8 decode + W8 lmhead + fp16 embed. Plane A: PASS 6/6 on S24.
flying101 changed pull request status to open
Adds the Hexagon v75 bundle under (source prism-ml/Ternary-Bonsai-1.7B-unpacked).
- Files: ternary-bonsai-1.7b-1024.json, tbonsai17b_decn1k_w8.bin, tbonsai17b_lmh_w8.bin, tbonsai17b_embed_f16.bin, tokenizer.json
- Precision: W8 decode + W8 lm-head + fp16 embed (Qwen3-style, decode-over-prompt seed)
- Compile: QAIRT 2.47.0, v75/soc 57; export gate 8/8 match HF greedy
- Device (Plane A, S24): 6/6 standardized prompts, coherent output
flying101 changed pull request title from Add Hexagon v75 (Galaxy S24 / SM8650, soc 57) bundle to v75 (S24/SM8650, soc 57): Ternary-Bonsai-1.7B W8 bundle — Plane A 6/6
v75 bundle details: Source prism-ml/Ternary-Bonsai-1.7B-unpacked. W8 decode + W8 lm-head + fp16 embed, Qwen3-style decode-over-prompt seed. QAIRT 2.47, v75/soc 57, export 8/8. Plane A on S24: 6/6 prompts, coherent think output.
sanmonga22 changed pull request status to merged