Skip to content

Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

C++

dsv41_fit: remove a duplicated constants block; one unit (tokens a second)

Niko122136a94857a10d
diff --git a/strata_control/dsv41_fit.py b/strata_control/dsv41_fit.py
index fea934cdf..96b841123 100644
--- a/strata_control/dsv41_fit.py
+++ b/strata_control/dsv41_fit.py
@@ -13,7 +13,7 @@ card, and computes the rest of the experts on the CPU, reading them in place fro
Measured (r9700, 4 x Radeon AI PRO R9700 32 GB, 251 GB RAM, 2026-10-20, f-dsv41-port 01389205e): per card 3.4-4.7 GB of dense
weights, 28.5-29.8 GB of experts (1,505-1,575 of its 3,840), ~0.6 GB free; 6,197 experts (116.5 GB) in graphics memory and
-9,163 (172.3 GB) on the CPU lane; --tp 2: prompts 143-181 tokens/s (3K prompts), answers 25-31 tokens/s. Pure functions, no I/O:
+9,163 (172.3 GB) on the CPU lane; --tp 2: prompts 143-181 tokens a second (3K prompts), answers 25-31 tokens a second. Pure functions, no I/O:
setup.py, tools/strata_mcp.py and strata_cli/launch import it. Recommend, never force: nothing here refuses a computer.
"""
@@ -44,16 +44,16 @@ REST_RAM_GB = 9.0 # RAM the engine and the system keep besid
SPARE_FITS_GB = 4.0 # RAM left beyond the CPU lane's experts below which "fits" becomes "tight"
DISK_HEADROOM_GB = 6.0 # the pack folder (hash tables, tokenizer) and room to work
SMALLEST_CARD_GB = 8.0 # a card under this adds little: the dense weights alone are ~4 GB of it per card of four
-MEASURED = ("4 x Radeon AI PRO R9700 (32 GB) with 251 GB of RAM: prompts of 3,000 tokens read at 143-181 tokens/s, answers "
- "at 25-31 tokens/s (--tp 2); one answer at a time")
+MEASURED = ("4 x Radeon AI PRO R9700 (32 GB) with 251 GB of RAM: prompts of 3,000 tokens read at 143-181 tokens a second, answers "
+ "at 25-31 tokens a second (--tp 2); one answer at a time")
# The two official IST-DAS sets (helcig/tmp-oct26-2, setup.py DSV41_ISTA_FILES): 4 shards each (the experts as IQ1_M / IQ2 / IQ3 types, the
# dense weights Q8_0 / Q5_K), plus ONE Engram (110.6 GB, IQ4_NL, shared by both sets, stays on the SSD) and the DSpark drafter (8.0 GB).
# Measured 2026-10-23 (f-dsv41-rel): quality against the original model (ppl +16% for 2.00, +7.7-8.3% for 2.25; code / agent +3.5% / +2.5%
-# on 2.25); speed on the owner PC (RTX 5070 12 GB, 64 GB RAM, 3.6 GB/s NVMe): answers ~4.7 tokens/s, prompts 121 tokens/s at 4K and
-# 231 tokens/s at 8K; RTX 3060 12 GB with 46 GB of RAM (the engine held to 36 GB): answers ~7.2 tokens/s on 2.00; 4 x Radeon AI PRO
-# R9700: answers ~31 tokens/s, prompts 330-373 tokens/s at 8K.
+# on 2.25); speed on the owner PC (RTX 5070 12 GB, 64 GB RAM, 3.6 GB/s NVMe): answers ~4.7 tokens a second, prompts 121 tokens a second at 4K and
+# 231 tokens a second at 8K; RTX 3060 12 GB with 46 GB of RAM (the engine held to 36 GB): answers ~7.2 tokens a second on 2.00; 4 x Radeon AI PRO
+# R9700: answers ~31 tokens a second, prompts 330-373 tokens a second at 8K.
ISTA_ENGRAM_BYTES = 110_595_290_720
ISTA_DSPARK_BYTES = 7_976_914_240
ISTA_META_BYTES = 5_780_589 # the first shard of Unsloth's UD-Q8_K_XL: only the metadata --engram-meta reads
@@ -68,9 +68,9 @@ ISTA_SETS = {
}
ISTA_DENSE_GB = 6.16 # the dense weights in the shards (attention, router, shared experts, the head; the embedding stays on the CPU)
ISTA_DENSE_ON_GPU_GB = 4.5 # what a card holds of them with a 12 GB card (measured on the 5070 / 3060: the engine places the rest itself)
-ISTA_MEASURED = ("an RTX 5070 (12 GB) with 64 GB of RAM and a 3.6 GB/s NVMe drive: answers about 4.7 tokens/s, a 4,000-token prompt "
- "read at 121 tokens/s; an RTX 3060 (12 GB) with 46 GB of RAM: about 7.2 tokens/s; four Radeon AI PRO R9700 cards: "
- "about 31 tokens/s")
+ISTA_MEASURED = ("an RTX 5070 (12 GB) with 64 GB of RAM and a 3.6 GB/s NVMe drive: answers about 4.7 tokens a second, a 4,000-token prompt "
+ "read at 121 tokens a second; an RTX 3060 (12 GB) with 46 GB of RAM: about 7.2 tokens a second; four Radeon AI PRO R9700 cards: "
+ "about 31 tokens a second")
def variant(model: str | None) -> dict | None:
@@ -89,41 +89,6 @@ def ista_download_bytes(model: str, have_engram: bool = False) -> int:
return v["set_bytes"] + ISTA_DSPARK_BYTES + ISTA_META_BYTES + (0 if have_engram else ISTA_ENGRAM_BYTES)
-# The two official IST-DAS sets (helcig/tmp-oct26-2, setup.py DSV41_ISTA_FILES): 4 shards each (the experts as IQ1_M / IQ2 / IQ3 types, the
-# dense weights Q8_0 / Q5_K), plus ONE Engram (110.6 GB, IQ4_NL, shared by both sets, stays on the SSD) and the DSpark drafter (8.0 GB).
-# Measured 2026-10-23 (f-dsv41-rel): quality against the original model (ppl +16% for 2.00, +7.7-8.3% for 2.25; code / agent +3.5% / +2.5%
-# on 2.25); speed on the owner PC (RTX 5070 12 GB, 64 GB RAM, 3.6 GB/s NVMe): answers ~4.7 tokens/s, prompts 121 tokens/s at 4K and
-# 231 tokens/s at 8K; RTX 3060 12 GB with 46 GB of RAM (the engine held to 36 GB): answers ~7.2 tokens/s on 2.00; 4 x Radeon AI PRO
-# R9700: answers ~31 tokens/s, prompts 330-373 tokens/s at 8K.
-ISTA_ENGRAM_BYTES = 110_595_290_720
-ISTA_DSPARK_BYTES = 7_976_914_240
-ISTA_META_BYTES = 5_780_589 # the first shard of Unsloth's UD-Q8_K_XL: only the metadata --engram-meta reads
-ISTA_SETS = {
- "IST-DAS-2.00bpw": {"title": "IST-DAS 2.00 bpw", "set_bytes": 138_238_892_160, "experts_gb": 132.07, "profile": "profile-2.00.bin",
- "ram_ok_gb": 40.0, "disk_total": 257.0,
- "quality": "about 16% above the original model's perplexity (the 2-bit price)"},
- "IST-DAS-2.25bpw": {"title": "IST-DAS 2.25 bpw", "set_bytes": 155_225_823_328, "experts_gb": 149.06, "profile": "profile-2.25.bin",
- "ram_ok_gb": 44.0, "disk_total": 274.0,
- "quality": "about 8% above the original model's perplexity (code and agent work about 2.5-3.5% above), "
- "the better of the two, 17 GB more disk"},
-}
-ISTA_DENSE_GB = 6.16 # the dense weights in the shards (attention, router, shared experts, the head; the embedding stays on the CPU)
-ISTA_DENSE_ON_GPU_GB = 4.5 # what a card holds of them with a 12 GB card (the engine places the rest itself)
-ISTA_MEASURED = ("an RTX 5070 (12 GB) with 64 GB of RAM and a 3.6 GB/s NVMe drive: answers about 4.7 tokens/s, a 4,000-token prompt "
- "read at 121 tokens/s; an RTX 3060 (12 GB) with 46 GB of RAM: about 7.2 tokens/s; four Radeon AI PRO R9700 cards: "
- "about 31 tokens/s")
-
-
-def variant(model: str | None) -> dict | None:
- """The official-set entry for a model name (case-insensitive), None for UD-Q8_K_XL and anything else."""
- return ISTA_SETS.get(variant_key(model)) if variant_key(model) else None
-
-
-def ista_download_bytes(model: str, have_engram: bool = False) -> int:
- v = variant(model)
- return v["set_bytes"] + ISTA_DSPARK_BYTES + ISTA_META_BYTES + (0 if have_engram else ISTA_ENGRAM_BYTES)
-
-
def _round1(x: float) -> float:
return round(float(x) * 10) / 10
@@ -167,7 +132,7 @@ def _plan_ista(model: str, cards: list[dict] | None, ram_gb: float | None, disk_
vendor: str | None) -> dict[str, Any]:
"""plan() for an official IST-DAS set (ISTA_SETS): the same keys, the measured speeds in plain words. RAM classes: with RAM for every
expert the CPU lane keeps ("fits"), with 40 GB or more the engine's RAM store keeps the hot experts and reads the rest from the SSD
- ("tight": measured 4.7 tokens/s on 64 GB, 7.2 on 46 GB), under 40 GB it is untested and slower ("spills"). Never a refusal."""
+ ("tight": measured 4.7 tokens a second on 64 GB, 7.2 on 46 GB), under 40 GB it is untested and slower ("spills"). Never a refusal."""
v = variant(model)
keep, ignored = gpu_fit.usable_cards(gpu_fit.for_vendor(list(cards or []), vendor)[0]) if vendor else gpu_fit.usable_cards(list(cards or []))
keep = [c for c in keep if float(c.get("vram_gb") or 0) >= SMALLEST_CARD_GB]
diff --git a/tools/test_setup_dsv41.py b/tools/test_setup_dsv41.py
index da81c54c8..7ddb917f1 100644
--- a/tools/test_setup_dsv41.py
+++ b/tools/test_setup_dsv41.py
@@ -373,7 +373,7 @@ class OfficialSets(unittest.TestCase):
self.assertEqual(p["args"], [])
p = dsv41_fit.plan([RTX5070], 64.0, 100.0, model=ISTA)
text = " ".join(p["lines"]) + " ".join(p["warnings"])
- for needle in ("4.7", "7.2", "31 tokens/s", "free disk space", "only once" if False else "downloads it once"):
+ for needle in ("4.7", "7.2", "31 tokens a second", "free disk space", "only once" if False else "downloads it once"):
self.assertIn(needle, text)
self.assertFalse(p["disk_ok"])
p225 = dsv41_fit.plan([RTX5070], 64.0, None, model="IST-DAS-2.25bpw")