mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2026-10-16 21:01:14 +08:00
yyj
4882505c9a
* [docs]: add reproducible Kimi K2.5 PyPI LoRA tutorial * [docs]: highlight Kimi K2.5 and K2.6 LoRA support * Update README to simplify LoRA fine-tuning entry * Add updates section to README.md * [docs]: clarify clean Kimi setup and held-out evaluation * [docs]: explain pinned model downloads on offline training hosts * [docs]: isolate Kimi caches in the tutorial work directory * [docs]: simplify Kimi LoRA installation and training tutorial
21 lines
661 B
YAML
21 lines
661 B
YAML
# Pair with train.yaml or train-neko.yaml using the external Accelerate launcher.
|
|
# Eight GPUs; FSDP2 parameter offload is separate from activation recomputation.
|
|
compute_environment: LOCAL_MACHINE
|
|
distributed_type: FSDP
|
|
fsdp_config:
|
|
fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
|
|
fsdp_cpu_ram_efficient_loading: true
|
|
fsdp_offload_params: true
|
|
fsdp_reshard_after_forward: true
|
|
fsdp_state_dict_type: FULL_STATE_DICT
|
|
fsdp_sync_module_states: true
|
|
fsdp_transformer_layer_cls_to_wrap: DeepseekV3DecoderLayer
|
|
fsdp_version: 2
|
|
machine_rank: 0
|
|
mixed_precision: bf16
|
|
num_machines: 1
|
|
num_processes: 8
|
|
rdzv_backend: static
|
|
same_network: true
|
|
use_cpu: false
|