| .. |
|
deepseek-v4-flash
|
ae1285dfdf
fix: reduce gpu-memory-utilization to 0.75 for deepseek-v4-flash
|
3 месяцев назад |
|
deepseek-v4-flash-dspark
|
57d686b17b
fix(deepseek-v4): add NCCL ABI fix + instanttensor overlay image
|
3 месяцев назад |
|
gemma4-12b
|
5931869d69
refactor: DistributedBackend as Option — remove None_ variant
|
4 месяцев назад |
|
glm51
|
09841103d8
fix: default DistributedBackend to None_ (absent means single-node)
|
4 месяцев назад |
|
glm52
|
ef03b8c784
feat: auto-scale PP when pipeline_parallel_size == required_nodes
|
4 месяцев назад |
|
glm52-hybrid-fp8-mxfp4
|
9e28b49887
feat(recipes): add volumes support and GLM-5.2 Hybrid FP8-MXFP4 recipe
|
3 месяцев назад |
|
glm52-int4-int8
|
9e28b49887
feat(recipes): add volumes support and GLM-5.2 Hybrid FP8-MXFP4 recipe
|
3 месяцев назад |
|
glm52-nf3-hybrid-context
|
9e28b49887
feat(recipes): add volumes support and GLM-5.2 Hybrid FP8-MXFP4 recipe
|
3 месяцев назад |
|
glm52-nf3-hybrid-fast
|
9e28b49887
feat(recipes): add volumes support and GLM-5.2 Hybrid FP8-MXFP4 recipe
|
3 месяцев назад |
|
laguna-m1-nvfp4
|
5837205e68
feat: add MiMo-V2.5 Omni NVFP4 recipe (2× DGX Spark, Ray, 1M ctx)
|
4 месяцев назад |
|
mimo-v2.5-nvfp4
|
d478170365
fix: switch MiMo recipe to mp backend, remove Ray-specific env vars
|
4 месяцев назад |
|
minimax-m3
|
ef03b8c784
feat: auto-scale PP when pipeline_parallel_size == required_nodes
|
4 месяцев назад |
|
minimax-m3-mxfp8
|
227eb18915
feat: auto-set VLLM_LOGGING_COLOR=1 on head node (override via recipe [env])
|
4 месяцев назад |
|
minimax-m3-nvfp4
|
ef03b8c784
feat: auto-scale PP when pipeline_parallel_size == required_nodes
|
4 месяцев назад |
|
qwen-agentworld-35b
|
4b08863c18
chore: update qwen-agentworld-35b recipe with tool call args and model name
|
4 месяцев назад |
|
test-mods
|
5931869d69
refactor: DistributedBackend as Option — remove None_ variant
|
4 месяцев назад |
|
test-sleep
|
1fa2811301
feat: default recipe port is 30000
|
4 месяцев назад |
Alexander Korolev
9e28b49887
feat(recipes): add volumes support and GLM-5.2 Hybrid FP8-MXFP4 recipe