Nikolai Kolodziej

Results 4 issues of Nikolai Kolodziej

This change wraps the spawn internals to copy files from snapshot if needed. The approach has a few downsides, if in fact the executable needs to read other snapshot files,...

no-stale

This change would allow the user to include additional relations in the `user` object when querying. The type for the includable relations is a bit suboptimal, as intellisense will show...

### System Info `/info` Output: ```json { "model_id": "casperhansen/llama-3.3-70b-instruct-awq", "model_sha": "64d255621f40b42adaf6d1f32a47e1d4534c0f14", "model_pipeline_tag": "text-generation", "max_concurrent_requests": 128, "max_best_of": 2, "max_stop_sequences": 4, "max_input_tokens": 8191, "max_total_tokens": 8192, "validation_workers": 2, "max_client_batch_size": 4, "router": "text-generation-router", "version":...

Installs `bitsandbytes>=0.44.0` and adds the `bitsandbytes` value to the `QUANTIZATION` select. See [vLLM bnb docs](https://docs.vllm.ai/en/latest/quantization/bnb.html).