darkmaniac7
/

TokForge-AccelerationPack-Qwen35-Draft

Text Generation

speculative-decoding

Model card Files Files and versions

TokForge-AccelerationPack-Qwen35-Draft

484 MB

Ctrl+K

Ctrl+K

1 contributor

History: 12 commits

darkmaniac7's picture

Upload README.md with huggingface_hub

86c289c verified 9 days ago

.gitattributes

1.66 kB
fix: replace VL model with text-only abliterated (prevents spec decode crash) 16 days ago
README.md

1.94 kB
Upload README.md with huggingface_hub 9 days ago
config.json

342 Bytes
feat: abliterated 0.8B with fixed graph (no FusedLinearAttention ops, is_visual=false) 16 days ago
export_args.json

1.11 kB
feat: abliterated 0.8B with fixed graph (no FusedLinearAttention ops, is_visual=false) 16 days ago
llm.mnn

2.15 MB
xet

feat: abliterated 0.8B with fixed graph (no FusedLinearAttention ops, is_visual=false) 16 days ago
llm.mnn.json

5.34 MB
xet

feat: abliterated 0.8B with fixed graph (no FusedLinearAttention ops, is_visual=false) 16 days ago
llm.mnn.weight

470 MB
xet

feat: abliterated 0.8B with fixed graph (no FusedLinearAttention ops, is_visual=false) 16 days ago
llm_config.json

8.69 kB
fix: is_visual=false + original taobao graph (was broken 320KB graph). 80% acceptance, +49% on 9B. 16 days ago
tokenizer.txt

6.47 MB
Upload folder using huggingface_hub 16 days ago