Commit History

Автор SHA1 Съобщение Дата
  AlpinDale 00503b9fc1 feat: non-uniform quantization via `compressed-tensors` for llama преди 5 месеца
  AlpinDale 0429cb2229 fix: only create embeddings and lm_head when necessary for PP преди 5 месеца
  AlpinDale 5289c14b24 feat: Asymmetric Tensor Parallel (#594) преди 5 месеца
  AlpinDale 9d7beaa5b9 chore: separate kv_scale into k_scale and v_scale преди 5 месеца
  AlpinDale 497bf64942 chore: simplify pipeline parallel code in llama преди 5 месеца
  AlpinDale 0f4a9ee77b quantized lm_head (#582) преди 5 месеца
  AlpinDale ae04f57ec1 feat: Pipeline Parallel support (#581) преди 6 месеца
  AlpinDale c5d8028668 fix: no need to redefine supports_vision and supports_lora in model class преди 6 месеца
  AlpinDale 56e0b8223c chore: add base class for LoRA-supported models преди 6 месеца
  AlpinDale 690110a051 feat: bitsandbytes quantization преди 6 месеца
  AlpinDale ac79d115b3 add guards for prefix caching, fp8, chunked, etc преди 6 месеца
  AlpinDale f4ea11b982 feat: initial support for activation quantization преди 6 месеца
  AlpinDale c1ed789835 fix: typo in llama.py преди 6 месеца
  AlpinDale 656459fd84 make fp8_e4m3 work on nvidia преди 6 месеца
  AlpinDale 9e73559eba make use of batched rotary embedding kernels to support long context lora преди 6 месеца
  AlpinDale 2ecfa98da9 re-fix mistral nemo преди 6 месеца
  AlpinDale 50b7c13db0 refactor: attention selector (#552) преди 6 месеца
  AlpinDale 54a4cef647 add bias and tie word embedding support for llama преди 6 месеца
  AlpinDale 639e48e47d fix: mistral nemo преди 6 месеца
  AlpinDale b178ae4b4a chore: generalize linear_method to be quant_method (#540) преди 6 месеца
  AlpinDale e7b1368156 feat: Phi3 support преди 7 месеца
  AlpinDale fca911ee0a vLLM Upstream Sync (#526) преди 7 месеца
  AlpinDale 9d81716bfd [v0.5.3] Release Candidate (#388) преди 9 месеца
  AlpinDale f8dfac6372 chore: attention refactor and upstream sync apr01 (#365) преди 10 месеца
  sgsdxzy 6ebac34dc1 chore: cleaner pre-llamafied Yi implementation (#352) преди 10 месеца
  AlpinDale 681e94611f fix: restore backwards compatibility with old Yi models (#351) преди 10 месеца
  AlpinDale da223153c6 feat&fix: cohere support and missing GPU blocks (#333) преди 11 месеца
  AlpinDale e42a78381a feat: switch from pylint to ruff (#322) преди 11 месеца
  AlpinDale 9810daa699 feat: INT8 KV Cache (#298) преди 11 месеца
  AlpinDale e31c6f0b45 feat: refactor modeling logic and support more models (#274) преди 11 месеца