CoolFace
Modelpublic

Jaidchen/Focus

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face
0likes78downloads
Model Card

<center> <img style='height: 16em' src='https://cdn-uploads.huggingface.co/production/uploads/63e9a1c4ccae1fe5c622b9f6/ADYJx3gsTmTVPuhHGMM5S.webp'/> </center>

Focus

repackaged Qwen 3.6 27B with a simplified architecture and minor opinionated improvements

  • —all vision-related components removed
  • —reduced storage and memory, faster inference
  • —zero loss of output quality

comparison

<table> <tr> <th></th> <th>Qwen 3.6 27B</th> <th>Focus</th> </tr> <tr> <td>author</td> <td><a href='https://huggingface.co/Qwen'>Alibaba Qwen</a></td> <td><a href='https://huggingface.co/Jaidchen'>Jaid</a></td> </tr> <tr> <td>repository</td> <td><a href='https://huggingface.co/Qwen/Qwen3.6-27B'>Qwen/Qwen3.6-27B</a></td> <td><a href='https://huggingface.co/Jaidchen/Focus'>Jaidchen/Focus</a></td> </tr> <tr> <td>architecture</td> <td><code>qwen35</code></td> <td><code>qwen35text</code></td> </tr> <tr> <td>Transformers handler</td> <td> <a href='https://huggingface.co/docs/transformers/modeldoc/qwen35#transformers.Qwen35ForConditionalGeneration'><code>Qwen35ForConditionalGeneration</code></a> </td> <td> <a href='https://huggingface.co/docs/transformers/modeldoc/qwen35#transformers.Qwen35ForCausalLM'><code>Qwen35ForCausalLM</code></a> </td> </tr> <tr> <td>tensor entries</td> <td>1199</td> <td>866</td> </tr> <tr style='opacity: 50%'> <td>tensor type</td> <td>bf16</td> <td>bf16</td> </tr> <tr> <td>parameters</td> <td>27 781 427 952</td> <td>27 320 697 856</td> </tr> <tr style='opacity: 50%'> <td>active</td> <td>100%</td> <td>100%</td> </tr> <tr style='opacity: 50%'> <td>vocabulary size</td> <td>248 320</td> <td>248 320</td> </tr> <tr style='opacity: 50%'> <td>context size</td> <td>262 144</td> <td>262 144</td> </tr> <tr style='opacity: 50%'> <td>MTP</td> <td>integrated</td> <td>integrated</td> </tr> <tr> <td>sampling strategy</td> <td>random sampling</td> <td>greedy/deterministic</td> </tr> <tr> <td>sampling parameters</td> <td> <div style='font-family: Jaidevka Code, JetBrains Mono, monospace; line-height: initial'> <span style='color: hsl(from currentColor 0 100% l)'>dosample</span>: <span style='color: hsl(from currentColor 50 80% l)'>true</span><br> <span style='color: hsl(from currentColor 0 100% l)'>temperature</span>: <span style='color: hsl(from currentColor 50 80% l)'>0.6</span><br> <span style='color: hsl(from currentColor 0 100% l)'>topk</span>: <span style='color: hsl(from currentColor 50 80% l)'>20</span><br> <span style='color: hsl(from currentColor 0 100% l)'>topp</span>: <span style='color: hsl(from currentColor 50 80% l)'>0.95</span></div> </td> <td> <div style='font-family: Jaidevka Code, JetBrains Mono, monospace; line-height: initial'> <span style='color: hsl(from currentColor 0 100% l)'>dosample</span>: <span style='color: hsl(from currentColor 50 80% l)'>false</span><br> <span style='color: hsl(from currentColor 0 100% l)'>temperature</span>: <span style='color: hsl(from currentColor 50 80% l)'>0</span><br> <span style='color: hsl(from currentColor 0 100% l)'>topk</span>: <span style='color: hsl(from currentColor 50 80% l)'>1</span><br> <span style='color: hsl(from currentColor 0 100% l)'>topp</span>: <span style='color: hsl(from currentColor 50 80% l)'>1</span></div> </div> </td> </tr> <tr> <td>input modality</td> <td>text, image, video</td> <td>text</td> </tr> <tr> <td>repository size</td> <td>55 586 107 940</td> <td>54 659 211 447</td> </tr> <tr> <td>model size</td> <td>55 562 855 904</td> <td>54 641 395 712</td> </tr> <tr> <td>splits</td> <td>15</td> <td>none</td> </tr> <tr> <td>Jinja template</td> <td><a href='https://huggingface.co/Qwen/Qwen3.5-27B/blob/main/chattemplate.jinja'>Qwen original</a></td> <td><a href='https://huggingface.co/Qwen/Qwen3.5-27B/blob/main/chattemplate.jinja'>Qwen original</a><br>+ <a href='https://huggingface.co/unsloth/Qwen3.5-27B/blob/main/chattemplate.jinja'>Unsloth tweaks</a><br>+ <a href='https://huggingface.co/froggeric/Qwen-Fixed-Chat-Templates/blob/main/chat_template.jinja'>Froggeric tweaks</a><br>+ unimodality patch<br>+ further custom tweaks</td> </tr> </table>

pros

  • —reduced storage needs
  • —reduced loading time
  • —reduced VRAM occupancy, thus more room for context
  • —increased inference speed
  • —simplified architecture, unlocking some further potential for optimizing low-level procedures

cons

  • —legally blind
  • —Pictures and video frames can still be present in the context without crashing, but their contents are no longer interpreted by the model and won’t do anything else than waste space.
  • —If you occasionally rely on those capabilities, I suggest adding a consult tool to your harness that calls a vision-enabled subagent model like Gemini Flash or GPT.
  • —reduced compatibility
  • —The simplified architecture is handled by the Qwen3_5ForCausalLM class which may not be included in your inference engine. In this case you would need to ask your agent or integrate it yourself.
  • —Inference engines confirmed to have native support:
  • —SGLang ≥ v0.5.17 (PR #32401)
  • —vLLM ≥ v0.26.0 (PR #50210)
  • —The applied coercions may confuse your inference engine in case it has fixed expectations about the model’s architecture and thus lead to unpredictable behavior.

caveats

  • —model file not split, possibly causing issues if intended to be stored on an HDD from the previous century
  • —random sampling disabled by default, less suitable for long-form writing, entertainment and casual chat

Jinja template changes

chat.jinja is reproducibly built from an untouched upstream template plus the ordered patch stack in jinja_build/. jinja_build/build.ps1 applies the patches lexicographically and overwrites the final chat.jinja.

  • —base: Qwen/Qwen3.6-27B original
  • —commit 6a9e13bd6fc8f0983b9b99948120bc37f49c13e9
  • —SHA-256 E84F32A23FDDA27689F868AA4A1A5621F41133E51A48D7F3EFCBEA2839574259
  • —adopted tweaks from Unsloth Qwen3.6
  • —Added developer role as alias for system. (lines 45–57)
  • —Merged leading system/developer messages into a single policy message. (lines 45–57, lines 67–75)
  • —Allowed histories without a normal human query. (lines 76–86)
  • —Made mapping-argument rendering portable by avoiding |items. (lines 122–130)
  • —adopted tweaks from Froggeric version
  • —Added preserve_thinking option to retain historical reasoning. (line 8, lines 225–230)
  • —Retained mid-conversation system/developer messages as Qwen system turns. (lines 165–182)
  • —Allowed string-valued message.thinking as fallback for historical reasoning. (lines 183–197)
  • —Applied boundary-aware </think> parsing. (lines 198–223)
  • —Avoided synthesizing empty historical thinking blocks. (lines 225–230)
  • —Supported both wrapped and direct tool calls, but never null-wrapped. (lines 231–237)
  • —Preserved non-empty string tool arguments. (lines 261–280)
  • —Used direct message indexing for tool-response grouping instead of loop.previtem/loop.nextitem. (line 165, lines 286–318)
  • —custom tweaks
  • —Merged any number of leading system/developer messages, generalizing Unsloth's two-message merge. (lines 45–57)
  • —Applied boundary-aware </think> parsing but excluded Froggeric's malformed-tag recovery. (lines 198–223)
  • —Raised an error containing the offending role instead of Qwen's generic unknown-role error. (lines 143–144)
  • —Used direct message indexing for tool-response grouping without Froggeric's error-escalation state. (lines 286–318)
  • —Guarded undefined tools/tool_calls and handled non-mapping content items defensively.
  • —Adopted Qwen's proposed continue_final_message fix for partial assistant prefills. (lines 130–132)
  • —unimodality patch
  • —Removed Qwen's vision-token machinery and rendered image/video content as [image] and [video]. (lines 1–41)

license

Apache 2.0 – inherited from Qwen 3.6 27B