{"data":{"items":[{"id":"94f7d2da-75dc-4517-8ca1-e23d8a15c606","excerpt":"Unsloth Desktop integrated browser does not respond to mouse clicks on Mac OS Golden Gate — <!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**W","url":"https://github.com/unslothai/unsloth/issues/13629","role":"pain","weight":0.9188334,"occurredAt":"2026-10-13T16:08:07.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"problem_report","painScore":0.85,"sentiment":-1,"confidence":0.49666667,"matchedPatterns":["doesnt_work","product:llama"],"statement":"Scr olling does not work either **Expected behavior:** **Actual behavior:** **Model and operation involved:** ## Diagnostics and logs ### Desktop application Click **Copy Diagnostics** on the error, startup, or update screen and paste the…","title":"Unsloth Desktop integrated browser does not respond to mouse clicks on Mac OS Golden Gate","body":"<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [x] Unsloth desktop application\n- [ ] Unsloth web UI (`unsloth studio`)\n- [ ] Unsloth CLI\n- [ ] Python package or notebook\n- [ ] Colab or Kaggle\n\n**Operating system and version:**\n On Mac Mini M1 16 GB OS Golden Gate 27.2**GPU model(s) and accelerator backend:**\n\n<!-- Examples: NVIDIA CUDA, AMD ROCm, Intel XPU, Apple MLX, CPU. -->\n\n**Versions:**\n<!--\nFor Unsloth, copy the Unsloth, package, desktop, and llama.cpp versions from\nSettings → About when available.\n-->Unsloth Version\nv0.1.906-beta\nPackage Version\n2026.10.4\nDesktop App Version\n0.1.906-beta\nllama.cpp Version\nb11541-mix-2500203\n## What happened?\n\n**Steps to reproduce:**\n\n1. Open Unsloth docs in integrated browser\n2.Click any button, Button seems to activate but nothing happens\n3. Scr\n\n<img width=\"1936\" height=\"1058\" alt=\"Image\" src=\"https://github.com/user-attachments/assets/0f1859db-e9dd-4dea-85db-2d657caf4fe2\" />\n\n<img width=\"1936\" height=\"1058\" alt=\"Image\" src=\"https://github.com/user-attachments/assets/541e9c43-57dc-4ea9-b8ec-2026006a9a2c\" />\n\nolling does not work either\n\n**Expected behavior:**\n\n**Actual behavior:**\n\n**Model and operation involved:**\n<!--\nInclude the model ID or filename and whether this involved installation,\nstartup, download, training, inference, GGUF, diffusion, or export.\n-->\n\n## Diagnostics and logs\n\n<!--\nRemove API keys, Hugging Face tokens, passwords, cookies, private prompts,\nlocal paths you do not want to disclose, and other sensitive information.\nDo not upload your entire ~/.unsloth/studio directory: it can contain auth\nstate, databases, chats, datasets, and models.\n-->\n\n### Desktop application\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n```text\nPASTE COPY DIAGNOSTICS HERE\n```\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\n<!-- Add screenshots or anything else that may help. -->\n","authorHash":"9c8d4e9ccbe70147081ce0f6659ad7a4","rating":null,"offTopic":false},{"id":"c8cbe22b-ef48-447d-83bf-d90d6b4a5062","excerpt":"[Bug] Intel: dual Arc Pro B60 on Vulkan hits ErrorDeviceLost mid-generation, chat locks until eject and reload — ---\nname: Unsloth Studio Bug\nabout: Report a problem with the Unsloth Studio desktop app or web UI\ntitle: \"Dual Intel Arc B60 Timeout\"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\nSearch existing issues before submit","url":"https://github.com/unslothai/unsloth/issues/11453","role":"pain","weight":0.70709044,"occurredAt":"2026-09-21T12:38:55.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.5050646,"matchedPatterns":["source:github"],"statement":"--- name: Unsloth Studio Bug about: Report a problem with the Unsloth Studio desktop app or web UI title: \"Dual Intel Arc B60 Timeout\" labels: bug assignees: \"\" --- ## Environment **Where are you using Unsloth?** - [X] Unsloth desktop appl…","title":"[Bug] Intel: dual Arc Pro B60 on Vulkan hits ErrorDeviceLost mid-generation, chat locks until eject and reload","body":"---\nname: Unsloth Studio Bug\nabout: Report a problem with the Unsloth Studio desktop app or web UI\ntitle: \"Dual Intel Arc B60 Timeout\"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [X] Unsloth desktop application\n- [ ] Unsloth web UI (`unsloth studio`)\n- [ ] Unsloth CLI\n- [ ] Python package or notebook\n- [ ] Colab or Kaggle\n\n**Operating system and version:**\nWindows 11,  WSL2\n\n**GPU model(s) and accelerator backend:**\nDual Intel Arc Pro B60, \nVulkan Drivers, \nAMD Ryazen 9 9950X3D ( GPU Drivers disabled for testing) \n\n**Versions:**\nUnsloth Desktop : v0.1.811-betable.\nPackage : \n2026.9.7\n\n## What happened?\n\n**Steps to reproduce:**\n\n1. Start with 2 GPU - both show in System settings,  enabled with the following startup command:\nset \"ZE_AFFINITY_MASK=0,1\"\n\necho ==========================================\necho Unsloth Studio - Intel B60 Multi-GPU\necho ==========================================\necho ZE_AFFINITY_MASK=%ZE_AFFINITY_MASK%\necho.\necho Starting Unsloth Studio...\necho.\n\nstart \"\" \"C:\\Users\\snow\\AppData\\Local\\Unsloth Studio (Desktop)\\unsloth-studio.exe\"\n\n2. Using any large LLM that forces Load to split across 2 GPU's . \nExample: Qwen3.8-27B-GUFF UD-Q8-K-XK\n\n**Expected behavior:**\nPrompt: Write an interesting version of the classic space invaders game\n- Create a HTML version of space invaders\n\n\n**Actual behavior:**\nModels Errors and Windows GPU memory usage drops to Zero\nChat locked\nEject Model\nReload Model\nResume to complete Game Generation\n\n\n**Model and operation involved:**\nUsing any large LLM that forces Load to split across 2 GPU's . \nExample: Qwen3.8-27B-GUFF UD-Q8-K-XK\n\n## Diagnostics and logs\n{\"timestamp\": \"2026-09-21T12:07:58.178062Z\", \"level\": \"error\", \"event\": \"Error during GGUF streaming: llama-server returned 500: {\\\"error\\\":{\\\"code\\\":500,\\\"message\\\":\\\"got exception: vk::Queue::submit: ErrorDeviceLost\\\",\\\"type\\\":\\\"server_error\\\"}}\", \"exception\": \"Traceback (most recent call last):\\n  File \\\"C:\\\\Users\\\\snow\\\\.unsloth\\\\studio\\\\unsloth_studio\\\\Lib\\\\site-packages\\\\studio\\\\backend\\\\routes\\\\inference.py\\\", line 25624, in gguf_stream_chunks\\n    cumulative = next_task.result()\\n                 ^^^^^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\snow\\\\AppData\\\\Local\\\\Programs\\\\Python\\\\Python312\\\\Lib\\\\asyncio\\\\threads.py\\\", line 25, in to_thread\\n    return await loop.run_in_executor(None, func_call)\\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\snow\\\\AppData\\\\Local\\\\Programs\\\\Python\\\\Python312\\\\Lib\\\\concurrent\\\\futures\\\\thread.py\\\", line 58, in run\\n    result = self.fn(*self.args, **self.kwargs)\\n             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\snow\\\\.unsloth\\\\studio\\\\unsloth_studio\\\\Lib\\\\site-packages\\\\studio\\\\backend\\\\core\\\\inference\\\\llama_cpp.py\\\", line 34066, in generate_chat_completion\\n    with self._open_stream(url, payload, cancel_event) as (\\n  File \\\"C:\\\\Users\\\\snow\\\\AppData\\\\Local\\\\Programs\\\\Python\\\\Python312\\\\Lib\\\\contextlib.py\\\", line 137, in __enter__\\n    return next(self.gen)\\n           ^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\snow\\\\.unsloth\\\\studio\\\\unsloth_studio\\\\Lib\\\\site-packages\\\\studio\\\\backend\\\\core\\\\inference\\\\llama_cpp.py\\\", line 33371, in _open_stream\\n    raise RuntimeError(\\nRuntimeError: llama-server returned 500: {\\\"error\\\":{\\\"code\\\":500,\\\"message\\\":\\\"got exception: vk::Queue::submit: ErrorDeviceLost\\\",\\\"type\\\":\\\"server_error\\\"}}\"}\n| Traceback (most recent call last):\n|   File \"C:\\Users\\snow\\.unsloth\\studio\\unsloth_studio\\Lib\\site-packages\\studio\\backend\\routes\\inference.py\", line 25624, in gguf_stream_chunks\n|     cumulative = next_task.result()\n|                  ^^^^^^^^^^^^^^^^^^\n|   File \"C:\\Users\\snow\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\asyncio\\threads.py\", line 25, in to_thread\n|     return await loop.run_in_executor(None, func_call)\n|            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n|   File \"C:\\Users\\snow\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\concurrent\\futures\\thread.py\", line 58, in run\n|     result = self.fn(*self.args, **self.kwargs)\n|              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n|   File \"C:\\Users\\snow\\.unsloth\\studio\\unsloth_studio\\Lib\\site-packages\\studio\\backend\\core\\inference\\llama_cpp.py\", line 34066, in generate_chat_completion\n|     with self._open_stream(url, payload, cancel_event) as (\n|   File \"C:\\Users\\snow\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 137, in __enter__\n|     return next(self.gen)\n|            ^^^^^^^^^^^^^^\n|   File \"C:\\Users\\snow\\.unsloth\\studio\\unsloth_studio\\Lib\\site-packages\\studio\\backend\\core\\inference\\llama_cpp.py\", line 33371, in _open_stream\n|     raise RuntimeError(\n| RuntimeError: llama-server returned 500: {\"error\":{\"code\":500,\"message\":\"got exception: vk::Queue::submit: ErrorDeviceLost\",\"type\":\"server_error\"}}\n\n### Desktop application\n\n<img width=\"877\" height=\"992\" alt=\"Image\" src=\"https://github.com/user-attachments/assets/cc723a34-254f-4b01-97ac-ba1752e5b038\" />\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n<img width=\"865\" height=\"757\" alt=\"Image\" src=\"https://github.com/user-attachments/assets/ba5b6959-c16d-4b92-97a1-d0d72eb88dc0\" />\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\n<!-- Add screenshots or anything else that may help. -->\n","authorHash":"c4f9271806c62d67d5045e05f3dd3776","rating":null,"offTopic":true},{"id":"3af07afc-ca39-4566-b3a6-1359ecd5a6c4","excerpt":"[Bug] Text generation failure with MiniCPM-V-4.7-1B Vulkan backend — <!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsl","url":"https://github.com/unslothai/unsloth/issues/14987","role":"pain","weight":0.61908746,"occurredAt":"2026-10-23T13:10:45.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.44220534,"matchedPatterns":["product:llama"],"statement":"[Bug] Text generation failure with MiniCPM-V-4.7-1B Vulkan backend.","title":"[Bug] Text generation failure with MiniCPM-V-4.7-1B Vulkan backend","body":"<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [X] Unsloth desktop application\n- [X] Unsloth web UI (`unsloth studio`)\n- [ ] Unsloth CLI\n- [ ] Python package or notebook\n- [ ] Colab or Kaggle\n\n**Operating system and version:**\n\n**GPU model(s) and accelerator backend:**\n<!-- Examples: NVIDIA CUDA, AMD ROCm, Intel XPU, Apple MLX, CPU. -->\nIntel(R) Arc(TM) 140V GPU\n**Versions:**\n<!--\nFor Unsloth, copy the Unsloth, package, desktop, and llama.cpp versions from\nSettings → About when available.\n-->\n\n## What happened?\nFailed text generation with model MiniCPM-V-4.7-1B after inputing an image and asking for energy meter status\n**Steps to reproduce:**\n\n1. Download MiniCPM-V-4.7-1B\n2. Load the model\n3. Prompt anything\n\n**Expected behavior:**\n Text generation\n**Actual behavior:**\nIt does not matter if the prompt contains an image or not. After prompting, the first word comes out of the model then generation fails.\n\n[llama-1792599533-port-58772-try0.log](https://github.com/user-attachments/files/33872829/llama-1792599533-port-58772-try0.log)\n[server-20261021-181808-pid1856.log](https://github.com/user-attachments/files/33872827/server-20261021-181808-pid1856.log)\n[backend-backend-1792599487960-2-s01.log](https://github.com/user-attachments/files/33872828/backend-backend-1792599487960-2-s01.log)\n<img width=\"2396\" height=\"1382\" alt=\"Image\" src=\"https://github.com/user-attachments/assets/1c11f698-4b30-407f-ba59-e2bee35a5556\" />\n<img width=\"2396\" height=\"1382\" alt=\"Image\" src=\"https://github.com/user-attachments/assets/96810ac6-7ff1-4513-95bf-1c3708243cc8\" />\n\n**Model and operation involved:**\n<!--\nInclude the model ID or filename and whether this involved installation,\nstartup, download, training, inference, GGUF, diffusion, or export.\n-->\n\n## Diagnostics and logs\n\n<!--\nRemove API keys, Hugging Face tokens, passwords, cookies, private prompts,\nlocal paths you do not want to disclose, and other sensitive information.\nDo not upload your entire ~/.unsloth/studio directory: it can contain auth\nstate, databases, chats, datasets, and models.\n-->\n\n### Desktop application\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n```text\nPASTE COPY DIAGNOSTICS HERE\n```\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\n<!-- Add screenshots or anything else that may help. -->\n","authorHash":"db741c4733cc51f6dd76f5600122aa43","rating":null,"offTopic":true},{"id":"56e5d13d-ab0e-4e63-9d6f-69509e0b5c36","excerpt":"False positive scan — <!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [x] Unsloth desktop application\n- [ ]","url":"https://github.com/unslothai/unsloth/issues/14306","role":"pain","weight":0.57554376,"occurredAt":"2026-10-20T02:01:43.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.41110268,"matchedPatterns":["product:llama"],"statement":"**Expected behavior:** **Actual behavior:** **Model and operation involved:** ## Diagnostics and logs ### Desktop application Click **Copy Diagnostics** on the error, startup, or update screen and paste the result below: If **Copy Diagnost…","title":"False positive scan","body":"<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [x] Unsloth desktop application\n- [ ] Unsloth web UI (`unsloth studio`)\n- [ ] Unsloth CLI\n- [ ] Python package or notebook\n- [ ] Colab or Kaggle\n\n**Operating system and version:**\n\n**GPU model(s) and accelerator backend:**\n<!-- Examples: NVIDIA CUDA, AMD ROCm, Intel XPU, Apple MLX, CPU. -->\n\n**Versions:**\n<!--\nFor Unsloth, copy the Unsloth, package, desktop, and llama.cpp versions from\nSettings → About when available.\n-->\n\n## What happened?\n\n**Steps to reproduce:**\n\n1.\n2.\n3.\n\n**Expected behavior:**\n\n**Actual behavior:**\n\n**Model and operation involved:**\n<!--\nInclude the model ID or filename and whether this involved installation,\nstartup, download, training, inference, GGUF, diffusion, or export.\n-->\n\n## Diagnostics and logs\n\n<!--\nRemove API keys, Hugging Face tokens, passwords, cookies, private prompts,\nlocal paths you do not want to disclose, and other sensitive information.\nDo not upload your entire ~/.unsloth/studio directory: it can contain auth\nstate, databases, chats, datasets, and models.\n-->\n\n### Desktop application\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n```text\nPASTE COPY DIAGNOSTICS HERE\n```\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\n<!-- Add screenshots or anything else that may help. -->\n","authorHash":"c8c3bdcec4fef526ab5ad96c211c3d5d","rating":null,"offTopic":true},{"id":"87376c97-fd9e-40f4-bd6d-d4e7ed468112","excerpt":"[Bug] Unsupported durable request fields: sandbox_level — <!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [","url":"https://github.com/unslothai/unsloth/issues/13010","role":"pain","weight":0.532,"occurredAt":"2026-10-08T02:23:40.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.38,"matchedPatterns":["product:llama"],"statement":"**Expected behavior:** **Actual behavior:** **Model and operation involved:** ## Diagnostics and logs ### Desktop application Click **Copy Diagnostics** on the error, startup, or update screen and paste the result below: If **Copy Diagnost…","title":"[Bug] Unsupported durable request fields: sandbox_level","body":"<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [ ] Unsloth desktop application\n- [ ] Unsloth web UI (`unsloth studio`)\n- [ ] Unsloth CLI\n- [ ] Python package or notebook\n- [ ] Colab or Kaggle\n\n**Operating system and version:**\n\n**GPU model(s) and accelerator backend:**\n<!-- Examples: NVIDIA CUDA, AMD ROCm, Intel XPU, Apple MLX, CPU. -->\n\n**Versions:**\n<!--\nFor Unsloth, copy the Unsloth, package, desktop, and llama.cpp versions from\nSettings → About when available.\n-->\n\n## What happened?\n\n**Steps to reproduce:**\n\n1.\n2.\n3.\n\n**Expected behavior:**\n\n**Actual behavior:**\n\n**Model and operation involved:**\n<!--\nInclude the model ID or filename and whether this involved installation,\nstartup, download, training, inference, GGUF, diffusion, or export.\n-->\n\n## Diagnostics and logs\n\n<!--\nRemove API keys, Hugging Face tokens, passwords, cookies, private prompts,\nlocal paths you do not want to disclose, and other sensitive information.\nDo not upload your entire ~/.unsloth/studio directory: it can contain auth\nstate, databases, chats, datasets, and models.\n-->\n\n### Desktop application\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n```text\nPASTE COPY DIAGNOSTICS HERE\n```\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\n<!-- Add screenshots or anything else that may help. -->\n","authorHash":"ef2b53e5f5412c931ccc3a02db306e92","rating":null,"offTopic":true},{"id":"bb4c35b3-62f9-4bab-862a-d2c36ef214ba","excerpt":"[Bug] Starting unsloth desktop clears bash history — ---\nname: Unsloth Studio Bug\nabout: Report a problem with the Unsloth Studio desktop app or web UI\ntitle: \"[Unsloth Bug] \"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging F","url":"https://github.com/unslothai/unsloth/issues/12678","role":"pain","weight":0.532,"occurredAt":"2026-10-04T10:21:46.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.38,"matchedPatterns":["product:llama"],"statement":"--- name: Unsloth Studio Bug about: Report a problem with the Unsloth Studio desktop app or web UI title: \"[Unsloth Bug] \" labels: bug assignees: \"\" --- ## Environment **Where are you using Unsloth?** - [x] Unsloth desktop application - […","title":"[Bug] Starting unsloth desktop clears bash history","body":"---\nname: Unsloth Studio Bug\nabout: Report a problem with the Unsloth Studio desktop app or web UI\ntitle: \"[Unsloth Bug] \"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [x] Unsloth desktop application\n- [ ] Unsloth web UI (`unsloth studio`)\n- [ ] Unsloth CLI\n- [ ] Python package or notebook\n- [ ] Colab or Kaggle\n\n**Operating system and version:** MacOS 27.0.1\n\n**GPU model(s) and accelerator backend:** Apple\n<!-- Examples: NVIDIA CUDA, AMD ROCm, Intel XPU, Apple MLX, CPU. -->\n\n**Versions:** 0.1.902-beta\n<!--\nFor Unsloth, copy the Unsloth, package, desktop, and llama.cpp versions from\nSettings → About when available.\n-->\n\n## What happened?\n\n**Steps to reproduce:**\n\n1. Open terminal and run `history` to see your history\n2. Run the desktop app\n3. Run `history` again and see it empty\n\n**Expected behavior:**\n\nMy history stays the same\n\n**Actual behavior:**\n\nHistory gone :(\n\n**Model and operation involved:**\n<!--\nInclude the model ID or filename and whether this involved installation,\nstartup, download, training, inference, GGUF, diffusion, or export.\n-->\n\n## Diagnostics and logs\n\n<!--\nRemove API keys, Hugging Face tokens, passwords, cookies, private prompts,\nlocal paths you do not want to disclose, and other sensitive information.\nDo not upload your entire ~/.unsloth/studio directory: it can contain auth\nstate, databases, chats, datasets, and models.\n-->\n\n### Desktop application\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n```text\nSession log: /Users/shagon/.unsloth/studio/logs/server/server-20261004-121415-pid23872.log\n{\"timestamp\": \"2026-10-04T10:14:15.778936Z\", \"level\": \"info\", \"event\": \"run_server startup begin api_only=True host=127.0.0.1 port=8888\"}\nChild-process cleanup on abnormal exit is NOT guaranteed: no kernel-level parent-death signal on this platform. Children may survive a crash or a force quit; the startup sweep reaps them on the next launch.\nLoading Unsloth Studio, please wait... (this can take a few minutes)\n  - loading PyTorch, Unsloth and Transformers...\n{\"timestamp\": \"2026-10-04T10:14:17.034905Z\", \"level\": \"info\", \"event\": \"Imported FastAPI app in 1233.2ms\"}\n  - Starting server...\n{\"timestamp\": \"2026-10-04T10:14:17.038145Z\", \"level\": \"info\", \"event\": \"Ensured Unsloth directories in 1262.7ms\"}\n[OK] Frontend loaded from /Users/shagon/.unsloth/studio/unsloth_studio/lib/python3.13/site-packages/studio/frontend/dist\n{\"timestamp\": \"2026-10-04T10:14:17.039689Z\", \"level\": \"info\", \"event\": \"run_server pre-uvicorn setup completed in 1264.3ms\"}\nINFO:     Started server process [23872]\nINFO:     Waiting for application startup.\n{\"timestamp\": \"2026-10-04T10:14:17.051393Z\", \"level\": \"info\", \"event\": \"Removing unsloth compiled cache: /Users/shagon/.unsloth/studio/compiled_cache\"}\n{\"timestamp\": \"2026-10-04T10:14:17.120490Z\", \"level\": \"info\", \"event\": \"lifespan pre-auth setup completed in 70.0ms\"}\n{\"timestamp\": \"2026-10-04T10:14:17.121937Z\", \"level\": \"info\", \"event\": \"lifespan startup completed in 71.5ms\"}\nINFO:     Application startup complete.\nINFO:     Unsloth Studio running on http://127.0.0.1:8888 (To stop: press Ctrl+C -- on macOS, Control+C not Command+C)\n{\"timestamp\": \"2026-10-04T10:14:17.127249Z\", \"level\": \"info\", \"event\": \"Uvicorn startup hook completed in 1351.8ms\"}\n{\"timestamp\": \"2026-10-04T10:14:17.127976Z\", \"level\": \"info\", \"event\": \"run_server uvicorn ready after 1352.6ms\"}\nTAURI_PORT=8888\n{\"timestamp\": \"2026-10-04T10:14:17.130650Z\", \"level\": \"info\", \"event\": \"parent_watchdog.started ppid=23831\"}\n\n🦥 Unsloth Studio is running\n────────────────────────────────────────────────────\n  On this machine -- open this in your browser:\n    http://127.0.0.1:8888\n    (same as http://localhost:8888)\n\n  API & health:\n    http://127.0.0.1:8888/api\n    http://127.0.0.1:8888/api/health\n────────────────────────────────────────────────────\n\n  Reachable on this machine only (bound to 127.0.0.1).\n  To expose it, turn on Settings > Remote & LAN > LAN access, or relaunch with:  unsloth studio -H 0.0.0.0 -p 8888\n  Only on trusted networks -- anyone who reaches this machine can use Unsloth.\nServer-side tools follow each request's enable_tools; the Unsloth UI's tool toggles decide. Pass --enable-tools to force them on for every request.\n\n  To stop Unsloth Studio: press Ctrl+C (Control+C, not Command+C, on macOS).\n────────────────────────────────────────────────────\n```\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\nhttps://github.com/user-attachments/assets/b5328b8c-cadf-4463-9c8f-2ab118fa1f1c","authorHash":"49847195795c24a2c5b7784af4d5cde4","rating":null,"offTopic":true},{"id":"8ef5ea84-ab8d-49ce-80f3-6336ea1fcec4","excerpt":"[Bug] stop generating buttons freeze. Keeps generating even after model unload and chat get freeze. — ---\nname: Unsloth Studio Bug\nabout: Report a problem with the Unsloth Studio web UI\ntitle: \"[Unsloth Bug] \"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\nSearch existing issues before submitting. Please do not remove the questio","url":"https://github.com/unslothai/unsloth/issues/12592","role":"pain","weight":0.532,"occurredAt":"2026-10-03T03:10:36.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.38,"matchedPatterns":["product:llama"],"statement":"--- name: Unsloth Studio Bug about: Report a problem with the Unsloth Studio web UI title: \"[Unsloth Bug] \" labels: bug assignees: \"\" --- ## Environment **Where are you using Unsloth?** - [ ] Unsloth desktop application - [x ] Unsloth web…","title":"[Bug] stop generating buttons freeze. Keeps generating even after model unload and chat get freeze.","body":"---\nname: Unsloth Studio Bug\nabout: Report a problem with the Unsloth Studio web UI\ntitle: \"[Unsloth Bug] \"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [ ] Unsloth desktop application\n- [x ] Unsloth web UI (`unsloth studio`)\n- [ ] Unsloth CLI\n- [ ] Python package or notebook\n- [ ] Colab or Kaggle\n\n**Operating system and version:**\n\n**GPU model(s) and accelerator backend:**\n<!-- Examples: NVIDIA CUDA, AMD ROCm, Intel XPU, Apple MLX, CPU. -->\n\n**Versions:**\n<!--\nFor Unsloth, copy the Unsloth, package, desktop, and llama.cpp versions from\nSettings → About when available.\n-->\n\n## What happened?\n\n**Steps to reproduce:**\n\n1.\n2.\n3.\n\n**Expected behavior:**\n\n**Actual behavior:**\n\n**Model and operation involved:**\n<!--\nInclude the model ID or filename and whether this involved installation,\nstartup, download, training, inference, GGUF, diffusion, or export.\n-->\n\n## Diagnostics and logs\n\n<!--\nRemove API keys, Hugging Face tokens, passwords, cookies, private prompts,\nlocal paths you do not want to disclose, and other sensitive information.\nDo not upload your entire ~/.unsloth/studio directory: it can contain auth\nstate, databases, chats, datasets, and models.\n-->\n\n### Desktop application\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n```text\nPASTE COPY DIAGNOSTICS HERE\n```\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\n<!-- Add screenshots or anything else that may help. -->\n","authorHash":"ef2b53e5f5412c931ccc3a02db306e92","rating":null,"offTopic":false},{"id":"5ba80f04-ef7a-4dd7-859f-599bd321f1fd","excerpt":"[Bug] Bonsai models suddenly do not load at all. — ---\nname: Bonsai models do not load\nabout: Ternary and 1bit models of prismml bonsai models used to launch either all in vram or split but now im getting errors \ntitle: \"[Unsloth Bug] \"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\nSearch existing issues before submitting. Pleas","url":"https://github.com/unslothai/unsloth/issues/11259","role":"pain","weight":0.532,"occurredAt":"2026-09-18T09:52:06.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"other","painScore":0.4,"sentiment":-1,"confidence":0.38,"matchedPatterns":["source:github"],"statement":"--- name: Bonsai models do not load about: Ternary and 1bit models of prismml bonsai models used to launch either all in vram or split but now im getting errors title: \"[Unsloth Bug] \" labels: bug assignees: \"\" --- ## Environment **Where a…","title":"[Bug] Bonsai models suddenly do not load at all.","body":"---\nname: Bonsai models do not load\nabout: Ternary and 1bit models of prismml bonsai models used to launch either all in vram or split but now im getting errors \ntitle: \"[Unsloth Bug] \"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\nSearch existing issues before submitting. Please do not remove the questions.\nNever post API keys, Hugging Face tokens, passwords, cookies, private prompts,\ndatasets, or other sensitive information.\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n- [X] Unsloth desktop application\n- [ ] Unsloth web UI (`unsloth studio`)\n- [ ] Unsloth CLI\n- [ ] Python package or notebook\n- [ ] Colab or Kaggle\n\n**Operating system and version:  Kernel: Linux 7.2.2-1-cachyos**\n\n**GPU model(s) and accelerator backend:**\nCPU: AMD Ryzen 7 8845HS (16) @ 5.10 GHz\nGPU 1: NVIDIA GeForce RTX 4070 Max-Q / Mobile [Discrete]\nCUDA\n\n**Versions:**\n<!--\nUnsloth Version v0.1.810-beta\nPackage Version 2026.9.6\nDesktop App Version 0.1.810-beta\nllama.cpp Version b11027-mix-3e83366 \n-->\n\n## What happened?\n\nFor me there were like 5 updates of either unsloth or llamacpp last few days, before that ive used to run ternary models of bonsai 27b without problems on my 8gbvram and 32gb ram with 64k context\n\n**Steps to reproduce:**\n\n1. update unsloth to latest\n2. download bonsai 27b ternary q4_1\n3. read the unexpected error\n\n**Expected behavior:**\nTernary loads to vram or ram and goes brrrr\n**Actual behavior:**\nIt does not and throws an error:\n\nFailed to load model: llama-server failed to start. Check that the GGUF file is valid and you have enough memory.\n\nllama-server output:\nerver: -----------------\n0.00.091.462 I srv    load_model: loading model '/home/virtualpierogi/.cache/huggingface/hub/models--prism-ml--Ternary-Bonsai-27B-gguf/snapshots/86e89f34c93201c3dfd5e5880fedb0022fc7e34d/Ternary-Bonsai-27B-dspark-Q4_1.gguf'\n0.00.108.547 E gguf_init_from_reader: tensor 'dspark.fc.weight' has offset 337718592, expected 357584192\n0.00.108.555 E gguf_init_from_reader: failed to read tensor data\n0.00.109.798 E llama_model_load: error loading model: llama_model_loader: failed to load model from /home/virtualpierogi/.cache/huggingface/hub/models--prism-ml--Ternary-Bonsai-27B-gguf/snapshots/86e89f34c93201c3dfd5e5880fedb0022fc7e34d/Ternary-Bonsai-27B-dspark-Q4_1.gguf\n0.00.109.802 E llama_model_load_from_file_impl: failed to load model\n0.00.109.828 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model\n0.00.109.870 E gguf_init_from_reader: tensor 'dspark.fc.weight' has offset 337718592, expected 357584192\n0.00.109.871 E gguf_init_from_reader: failed to read tensor data\n0.00.109.877 E llama_model_load: error loading model: llama_model_loader: failed to load model from /home/virtualpierogi/.cache/huggingface/hub/models--prism-ml--Ternary-Bonsai-27B-gguf/snapshots/86e89f34c93201c3dfd5e5880fedb0022fc7e34d/Ternary-Bonsai-27B-dspark-Q4_1.gguf\n0.00.109.877 E llama_model_load_from_file_impl: failed to load model\n0.00.109.879 E cmn  common_init_: failed to load model '/home/virtualpierogi/.cache/huggingface/hub/models--prism-ml--Ternary-Bonsai-27B-gguf/snapshots/86e89f34c93201c3dfd5e5880fedb0022fc7e34d/Ternary-Bonsai-27B-dspark-Q4_1.gguf'\n0.00.109.882 E srv    load_model: failed to load model, '/home/virtualpierogi/.cache/huggingface/hub/models--prism-ml--Ternary-Bonsai-27B-gguf/snapshots/86e89f34c93201c3dfd5e5880fedb0022fc7e34d/Ternary-Bonsai-27B-dspark-Q4_1.gguf'\n0.00.109.884 I srv    operator(): operator(): cleaning up before exit...\n0.00.110.392 E srv  llama_server: exiting due to model loading error\n\nFull log: /home/virtualpierogi/.unsloth/studio/logs/llama-server/llama-1789724835-port-35519-try0.log\n**Model and operation involved:**\n<!--\nInclude the model ID or filename and whether this involved installation,\nstartup, download, training, inference, GGUF, diffusion, or export.\n-->\n\n## Diagnostics and logs\n\n<!--\nRemove API keys, Hugging Face tokens, passwords, cookies, private prompts,\nlocal paths you do not want to disclose, and other sensitive information.\nDo not upload your entire ~/.unsloth/studio directory: it can contain auth\nstate, databases, chats, datasets, and models.\n-->\n\n### Desktop application\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n```text\nPASTE COPY DIAGNOSTICS HERE\n```\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\n<!-- Add screenshots or anything else that may help. -->\nt","authorHash":"596bb3d2a336f9202330b8388aeff159","rating":null,"offTopic":false},{"id":"74700ee9-d502-4e7f-8dfd-145817bd6e46","excerpt":"[Bug] Getting error \"indices should be either on cpu or on the same device as the indexed tensor (cuda:0)\" when trying to train Gemma 4 31b since upgrading to v0.1.806-beta — ---\nname: Unsloth Studio Bug\nabout: Report a problem with the Unsloth Studio desktop app or web UI\ntitle: \"[Unsloth Bug] \"\nlabels: bug\nassignees:","url":"https://github.com/unslothai/unsloth/issues/11952","role":"pain","weight":0.49234784,"occurredAt":"2026-09-25T18:01:41.000Z","sourceKey":"github","sourceName":"GitHub","credibility":0.78,"venue":"unslothai/unsloth","intent":"other","painScore":0.29565218,"sentiment":-0.73913044,"confidence":0.38,"matchedPatterns":["product:mistral"],"statement":"--- name: Unsloth Studio Bug about: Report a problem with the Unsloth Studio desktop app or web UI title: \"[Unsloth Bug] \" labels: bug assignees: \"\" --- ## Environment **Where are you using Unsloth?** Unsloth desktop application **Operatin…","title":"[Bug] Getting error \"indices should be either on cpu or on the same device as the indexed tensor (cuda:0)\" when trying to train Gemma 4 31b since upgrading to v0.1.806-beta","body":"---\nname: Unsloth Studio Bug\nabout: Report a problem with the Unsloth Studio desktop app or web UI\ntitle: \"[Unsloth Bug] \"\nlabels: bug\nassignees: \"\"\n---\n\n<!--\n\n-->\n\n## Environment\n\n**Where are you using Unsloth?**\n\n Unsloth desktop application\n\n\n**Operating system and version:**\nWindows 11 \n\n**GPU model(s) and accelerator backend:**\n<!-- Examples: NVIDIA CUDA, AMD ROCm, Intel XPU, Apple MLX, CPU. -->\n4 x RTX 4060Ti 16gb  CUDA\n**Versions:**\nv0.1.814-beta\nPackage Version\n2026.9.10\nllama.cpp Version\nb11160-mix-a6922cc\n<!--\nFor Unsloth, copy the Unsloth, package, desktop, and llama.cpp versions from\nSettings → About when available.\n-->\n\n## What happened?\n\nI had been working on a CPT>SFT>DPO project with Gemma 4 31b for a couple of weeks and it been going smooth until I upgraded to v0.1.806-beta last week then i immediately started getting the above error. I tried rolling back but it didnt help. I can still train Gemma 4 12b, Nemo 12b, Mistral small 24 normally but gemma4 31b errors before first step\n\n**Steps to reproduce:**\n\n1.Start unsloth\n2.choose Gemma 4 31b and start training run\n\n\n**Expected behavior:**\nTraining runs as normal\n**Actual behavior:**\nGet error \"indices should be either on cpu or on the same device as the indexed tensor (cuda:0)\" \n\n[UnslothTraceback.txt](https://github.com/user-attachments/files/32665431/UnslothTraceback.txt)\n\n**Model and operation involved:**\n<!--\nInclude the model ID or filename and whether this involved installation,\nstartup, download, training, inference, GGUF, diffusion, or export.\n-->\n\n## Diagnostics and logs\n\n<!--\nRemove API keys, Hugging Face tokens, passwords, cookies, private prompts,\nlocal paths you do not want to disclose, and other sensitive information.\nDo not upload your entire ~/.unsloth/studio directory: it can contain auth\nstate, databases, chats, datasets, and models.\n-->\n\n### Desktop application\n\nClick **Copy Diagnostics** on the error, startup, or update screen and paste the\nresult below:\n\n```text\n==((====))==  Unsloth - 2x faster free finetuning | Num GPUs used = 2\n   \\\\   /|    Num examples = 10,282 | Num Epochs = 2 | Total steps = 644\nO^O/ \\_/ \\    Batch size per device = 2 | Gradient accumulation steps = 16\n\\        /    Data Parallel GPUs = 1 | Total batch size (2 x 16 x 1) = 32\n \"-____-\"     Trainable parameters = 122,429,440 of 31,395,515,952 (0.39% trained)\nUnsloth: Not an error, but Gemma4ForConditionalGeneration does not accept `num_items_in_batch`.\nUsing gradient accumulation will be very slightly less accurate.\nRead more on gradient accumulation issues here: https://unsloth.ai/blog/gradient\n{\"timestamp\": \"2026-09-25T16:51:16.353966Z\", \"level\": \"info\", \"event\": \"request_completed\", \"method\": \"GET\", \"path\": \"/api/train/runs\", \"status_code\": 200, \"process_time_ms\": 1343.38}\n{\"timestamp\": \"2026-09-25T16:51:16.897766Z\", \"level\": \"info\", \"event\": \"request_completed\", \"method\": \"GET\", \"path\": \"/api/inference/monitor\", \"status_code\": 200, \"process_time_ms\": 3.76}\n{\"timestamp\": \"2026-09-25T16:51:26.955645Z\", \"level\": \"info\", \"event\": \"request_completed\", \"method\": \"GET\", \"path\": \"/api/inference/monitor\", \"status_code\": 200, \"process_time_ms\": 4.02}\n{\"timestamp\": \"2026-09-25T16:51:31.314679Z\", \"level\": \"info\", \"event\": \"request_completed\", \"method\": \"GET\", \"path\": \"/api/train/runs\", \"status_code\": 200, \"process_time_ms\": 1302.4}\n{\"timestamp\": \"2026-09-25T16:51:31.825676Z\", \"level\": \"error\", \"event\": \"Training error: indices should be either on cpu or on the same device as the indexed tensor (cuda:0)\"}\n{\"timestamp\": \"2026-09-25T16:51:31.831675Z\", \"level\": \"error\", \"event\": \"Full traceback:\\nTraceback (most recent call last):\\n  File \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\unsloth_studio\\\\Lib\\\\site-packages\\\\studio\\\\backend\\\\core\\\\training\\\\trainer.py\\\", line 4537, in _train_worker\\n    self.trainer.train(\\n  File \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\unsloth_studio\\\\Lib\\\\site-packages\\\\unsloth\\\\trainer.py\\\", line 1549, in _train_with_reset\\n    return _orig_train(*train_args, **train_kwargs)\\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\compiled_cache\\\\UnslothSFTTrainer.py\\\", line 116, in wrapper\\n    output = f(self, *args, **kwargs)\\n             ^^^^^^^^^^^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\.venv_t5_550\\\\transformers\\\\trainer.py\\\", line 1424, in train\\n    return inner_training_loop(\\n           ^^^^^^^^^^^^^^^^^^^^\\n  File \\\"<string>\\\", line 81, in _fast_inner_training_loop\\n  File \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\.venv_t5_550\\\\transformers\\\\trainer.py\\\", line 1734, in _run_epoch\\n    tr_loss_step = self.training_step(model, inputs, num_items_in_batch)\\n                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\compiled_cache\\\\UnslothSFTTrainer.py\\\", line 1438, in training_step\\n    return super().training_step(*args, **kwargs)\\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\unsloth_studio\\\\Lib\\\\site-packages\\\\unsloth\\\\models\\\\_utils.py\\\", line 3838, in _unsloth_training_step_settling_fallbacks\\n    return _training_step_before_settle(self, *args, **kwargs)\\n           ^^^^^^^^^^^^^^\\n... [21877 chars omitted; raise UNSLOTH_STUDIO_MAX_EXCEPTION_CHARS to see it all] ...\\nile \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\unsloth_studio\\\\Lib\\\\site-packages\\\\torch\\\\_dynamo\\\\_trace_wrapped_higher_order_op.py\\\", line 100, in forward\\n    return torch.ops.aten.index(x, indices)\\n           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\\n  File \\\"C:\\\\Users\\\\mikee\\\\.unsloth\\\\studio\\\\unsloth_studio\\\\Lib\\\\site-packages\\\\torch\\\\_ops.py\\\", line 1269, in __call__\\n    return self._op(*args, **kwargs)\\n           ^^^^^^^^^^^^^^^^^^^^^^^^^\\nRuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cuda:0)\\n\"}\n{\"timestamp\": \"2026-09-25T16:51:31.832676Z\", \"level\": \"error\", \"event\": \"Training error: indices should be either on cpu or on the same device as the indexed tensor (cuda:0)\"}\n\n```\n\nIf **Copy Diagnostics** is unavailable, attach the newest relevant files from:\n\n- Windows: `%USERPROFILE%\\.unsloth\\studio\\`\n- Linux/macOS: `~/.unsloth/studio/`\n\nUseful files include:\n\n- `tauri.log` and, if relevant, `tauri.log.1`\n- `logs/install-*.log`, `logs/update-*.log`, `logs/repair-*.log`, or `logs/backend-*.log`\n- The newest `logs/server/server-*.log`\n\n### Unsloth web UI\n\nAttach the newest relevant items:\n\n- Linux/macOS/Windows: `~/.unsloth/studio/logs/server/server-*.log`\n  (use `%USERPROFILE%\\.unsloth\\studio\\logs\\server\\` on Windows)\n- Linux/macOS shortcut launches only: `~/.local/share/unsloth/studio.log`\n- Terminal output if Unsloth was launched from a terminal\n- Browser Console errors for browser-only problems\n\nIf you configured `UNSLOTH_STUDIO_HOME` or `STUDIO_HOME`, look under\n`<CUSTOM_STUDIO_HOME>/logs/` instead.\n\n### Model-specific logs, if applicable\n\n- GGUF/llama.cpp: newest `~/.unsloth/studio/logs/llama-server/*.log`\n- Diffusion GGUF: newest `~/.unsloth/studio/logs/diffusion-server/*.log`\n- Training resume/checkpoint bug: relevant `trainer_state.json`\n\n### Python package or notebook, if applicable\n\nPaste the complete traceback and minimal reproduction:\n\n```python\n# Remove all tokens and private data.\n```\n\nAlso include Python, Unsloth, unsloth_zoo, PyTorch, Transformers, and TRL\nversions, plus `nvidia-smi` output when applicable.\n\n## Additional context\n\n<!-- Add screenshots or anything else that may help. -->\n","authorHash":"8dd167e1293f70de68062c60dda9de3b","rating":null,"offTopic":true}],"breakdown":[{"sourceKey":"github","sourceName":"GitHub","count":9}],"total":9}}