# prain placement: prithivMLmods/gemma-4-26B-A4B-it-Uncensored-MAX on an 8 GB RX 6600 XT
# Model: 30 MoE layers x 128 experts, top-8 routing, 11.34 MiB per expert (BF16).
# VRAM, measured: 6.22 GiB weights (4.46 dense + 5 slots x 30 layers) + 0.75 GiB KV cache: 7.9 of
# 8 GiB in use. Serve with --attention-backend TRITON_ATTN (ROCM_ATTN needs more LDS than RDNA2 has).
# RAM tier: pinning is very slow behind a strict IOMMU (Thunderbolt eGPU: minutes per GiB), so this
# file keeps no RAM tier; the kernel page cache holds the hot part of the pack instead.
#
# Tiers:  vram = pinned in a VRAM slot for the whole run
#         ram  = pinned host memory, copied into a cache slot when a token needs it
#         disk = read from the pack file in pack_dir when a token needs it
#                (the kernel page cache still caches it; nothing is pinned)
# [layers."<selector>"] sections apply in file order; a later section wins per expert.
# Layer selectors: "*", "12", "0-9,20".   Expert lists: "*", "0-15,32", [1, 2, 3].
# Check what a file costs:   prain check <this file> <model config.json>
# Compare variants offline:  prain sim <trace> <this file> <model config.json>
#                            prain tune <trace> <model config.json> --expert-vram-gib 1.7 ...
# Relative paths start at this file's directory; "~" works too.

pack_dir = "../models/packs/gemma4-26b-a4b"  # 42.5 GiB, written on the first run
default_tier = "disk"                    # runs on any RAM size; `prain tune` does better
cache_slots = 5                          # 30 x 5 x 11.34 MiB = 1.66 GiB of VRAM
stats_path = "../models/packs/gemma4-26b-a4b/stats.json"
stats_every = 64                         # MoE forward passes between stats/trace writes
trace_path = "../models/packs/gemma4-26b-a4b/trace.jsonl.gz"  # routing trace for sim/tune

[policy]              # which expert a cache slot gives up
eviction = "lfu"      # lru | lfu (decayed frequency) | gdsf (frequency x refetch cost)
heat = "weight"       # what makes an expert hot: count (tokens) | weight (router mass) |
                      # recency (router mass, later tokens weigh more: good for warm-ups)
lfu_half_life = 64    # forwards until an LFU score halves
disk_cost = 4.0       # gdsf: a disk expert costs this many RAM experts to refetch
recency_half_life = 32  # heat = "recency": tokens until a token's weight halves
scan_tokens = 64      # a forward with this many tokens (a prompt) is a scan:
scan_keep = 0         #   it may not evict the scan_keep most valuable cached experts,
promote = 2           #   and only its `promote` hottest experts get full priority

[prefetch]            # copy experts before their layer runs, on a side stream
mode = "none"         # none | pregate (run layer L+distance's router on layer L's input)
distance = 1          # 1 or 2 layers ahead
budget = 2            # experts per layer and forward
min_prob = 0.05       # skip guesses below this router probability

[miss]                # an expert that is not in VRAM
mode = "auto"         # copy | cpu (always compute it on the CPU) | auto (CPU if it has
cpu_max_tokens = 2    #   at most cpu_max_tokens tokens, else copy). Measured on the Thunderbolt
                      #   eGPU host: a copy costs ~4.6 ms + disk read, the CPU ~2.5 ms per token
cpu_threads = 16      # CPU threads of the vLLM worker (sets OMP_NUM_THREADS)
admit = 0             # cpu/auto: copy this many CPU-run RAM-tier experts into VRAM afterwards

[serve]               # parallel requests (`prain serve` passes these to vLLM)
max_requests = 2      # requests vLLM runs at once (--max-num-seqs); warm-ups count too
kv_cache_gib = 0.75   # KV cache VRAM shared by them (--kv-cache-memory-bytes). vLLM logs
                      #   "Maximum concurrency for 4096 tokens per request: N.NNx" at start;
                      #   below max_requests, long requests wait or get preempted. Every GiB
                      #   here is 90 expert cache slots less.

[draft]               # answers that arrive while others use the GPU (`prain web --config`)
full_requests = 1     # answers that run with all their experts; later ones run at once as
                      #   drafts on the experts already in VRAM (= max_requests: never draft)
min_weight = 0.5      # a draft token still copies its heaviest missing experts until this
                      #   share of its router weight is computed (0 = only what is in VRAM)
refine = "regenerate" # regenerate: once a full slot frees, the draft is cut short and the
                      #   answer generated again with all experts | none: keep the draft