77 lines
2.8 KiB
Bash
77 lines
2.8 KiB
Bash
# Tunables for frank-default-small (Ornith-1.5-35B-A3B CRACK Q6_K, MTP MoE,
|
|
# Dankserver, RTX 3090).
|
|
#
|
|
# Runtime: upstream llama.cpp llama-server (b10121). RTX 3090 24 GB; Ryzen
|
|
# 7 7700X; 62 GB RAM. MoE expert weights pinned to CPU with an explicit
|
|
# --override-tensor regex (not --n-cpu-moe). Replaced the HauhauCS
|
|
# Qwen3.6-35B-A3B model on 2026-08-24. Switched from CRACK Q8_0 to Q6_K on
|
|
# 2026-08-25.
|
|
#
|
|
# WARNING: do not name any variable TEMP, HOST, or GPU_MAX_HEAP_SIZE. The
|
|
# ROCm HIP runtime reads those bare env vars (see the 2026-08-13 TEMP
|
|
# incident). This box is NVIDIA/CUDA, but the rule keeps the env portable.
|
|
|
|
# Model files (Q6_K CRACK GGUF + f16 mmproj vision encoder + MTP head bundled
|
|
# in the GGUF). Dealign2026-08-24.
|
|
MODEL_DIR=/home/dank/models/Ornith-1.5-35B-A3B-CRACK
|
|
MODEL_FILE=Ornith-1.5-35B-A3B-CRACK-Q6_K.gguf
|
|
MMPROJ_FILE=mmproj-Ornith-1.5-35B-A3B-f16.gguf
|
|
# Upstream llama.cpp binary (b10121).
|
|
LLAMA_SERVER=/home/dank/llama.cpp/build/bin/llama-server
|
|
|
|
# Qwen-Sharp chat template from peculiar-ragdoll (tracked in the repo,
|
|
# staged on the host).
|
|
# Source: https://huggingface.co/peculiar-ragdoll/Qwen-Sharp-Chat-Templates
|
|
CHAT_TEMPLATE_FILE=/home/dank/serverAppConfigs/scripts/chat-template.jinja
|
|
|
|
ALIAS=frank-default-small
|
|
BIND_ADDR=0.0.0.0
|
|
LISTEN_PORT=7770
|
|
|
|
# 3 parallel slots x 163840 ctx each (262144 lowered by 96k per stream for
|
|
# cache and checkpoint headroom) = 491520 total context pool. The per-slot
|
|
# 163840 will be advertised in models.json after validation.
|
|
CTX_SIZE=491520
|
|
PARALLEL=3
|
|
|
|
# Spec: MoE model with a bundled MTP head -> native draft-mtp, n_max 4
|
|
# (the knee). ngram under MTP hurts; MTP alone is the proven best for MoE.
|
|
SPEC_TYPE=draft-mtp
|
|
SPEC_DRAFT_N_MAX=4
|
|
SPEC_DRAFT_N_MIN=0
|
|
SPEC_DRAFT_P_MIN=0.0
|
|
|
|
# Expert offload via explicit tensor override (replaces --n-cpu-moe). The Q6_K
|
|
# layer audit found 630 MiB of expert tensors per block. Keep the selected
|
|
# full-attention blocks 27 and 31 plus output-adjacent blocks 33-40 on CUDA,
|
|
# including the bundled MTP block 40. Move all other expert tensors to CPU.
|
|
# All non-expert tensors stay on the 24 GB card. Buffer type "CPU" is the CPU
|
|
# buft name on this host.
|
|
OVERRIDE_EXPERT='blk\.(0|1|2|3|4|5|6|7|8|9|10|11|12|13|14|15|16|17|18|19|20|21|22|23|24|25|26|28|29|30|32)\.ffn_(up|down|gate|gate_up)_(ch|)exps'
|
|
OVERRIDE_BUFT=CPU
|
|
N_GPU_LAYERS=all
|
|
|
|
# KV cache on GPU, both K and V at f16.
|
|
CACHE_TYPE_K=f16
|
|
CACHE_TYPE_V=f16
|
|
|
|
VERBOSITY=4
|
|
# Store context checkpoints in RAM to hold peak GPU memory down.
|
|
CTX_CHECKPOINTS=16
|
|
|
|
# Big prefill batches.
|
|
BATCH_SIZE=4096
|
|
UBATCH_SIZE=2048
|
|
|
|
# Eager RAM load (expert weights on CPU; ~31 GB RAM).
|
|
NO_MMAP=1
|
|
|
|
# Sampling (Dealign Ornith model card, thinking-general).
|
|
TEMPERATURE=1.0
|
|
TOP_K=20
|
|
TOP_P=0.95
|
|
MIN_P=0.0
|
|
|
|
# Compute threads lowered to 8 so other processes keep a core.
|
|
THREADS=7
|
|
THREADS_HTTP=8 |