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++

setup: repin the DeepSeek 2.00/2.25 sets to ISTA-DASLab's final files (flat repository 0d8f49fb, IQ2_XXS = 2.00 bpw, IQ2_XS = 2.25 bpw, DeepSeek-V4.1-Flash-Engram-IQ4_NL.gguf); the first release's local files (model-2.00bpw-..., engram-IQ4_NL.gguf, in 2.00bpw/ folders) stay accepted by their old size and SHA-256 and are never downloaded again; tests

Niko1221598372c3fd20
diff --git a/setup.py b/setup.py
index c5f2eb5c4..46de4599b 100644
--- a/setup.py
+++ b/setup.py
@@ -99,7 +99,7 @@ HF_REVISIONS = {
"unsloth/Qwen3.6-35B-A3B-MTP-GGUF": "5bc3e238d916f48a861bac2f8a1990a0e9b7e98d", # 2026-10-12
"agentionai/Qwen3.8-Flash-Next-Gyro-GGUF": "1ada08422e404aca51ab9df665f089f34c419c6b", # 2026-10-13
"unsloth/DeepSeek-V4.1-Flash-1017-GGUF": "f924b94f19c1fea02c06a65c8475b0a45e2b88b1", # 2026-10-20 (the port's reference)
- "helcig/tmp-oct26-2": "851709dcf0d97f091a4bdce758b2e4b0b468dd26", # 2026-10-23 (the official IST-DAS DeepSeek sets)
+ "ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF": "0d8f49fb15df73789a9c237a1b8b7b088423dbc0", # 2026-10-24 (ISTA-DASLab's final 2.00 / 2.25 bpw sets and Engram)
}
@@ -678,15 +678,31 @@ DSV41_FILES = {
"dspark-DeepSeek-V4.1-Flash-1017-Q8_0.gguf": (7976914240, "6ba6ac1b331c540691a209e7fc97dec2a24e295acb1e755c84c7d24a4373927a"),
"mmproj-BF16.gguf": (972208288, "9a1ef7c40d0d702c7ed8c5f863f60abae64091387ba3fce0ad19ce9beeda3bf9"),
}
-# The two OFFICIAL IST-DAS sets (helcig/tmp-oct26-2 at the revision above; Hub LFS sha256 = what the files hash to): 4 shards each, plus ONE
-# Engram table (110.6 GB, the same file for both sets: downloaded once into the shared folder) and the DSpark drafter and the 5.8 MB
-# metadata shard of Unsloth's set that pack_gguf --engram-meta reads (both pinned in DSV41_FILES above). This is the set a normal PC runs
-# (64 GB RAM + a 12 GB card; docs/DEEPSEEK41.md); the UD-Q8_K_XL entry stays for the big machines.
-DSV41_ISTA_REPO = "helcig/tmp-oct26-2"
-DSV41_ISTA_ENGRAM = "engram-IQ4_NL.gguf"
-DSV41_ISTA_SHARD = "model-{bpw}-{i:05d}-of-00004.gguf"
+# The two OFFICIAL IST-DAS sets (ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF at the revision above, a flat repository; Hub LFS sha256 = what the
+# files hash to): IQ2_XXS is the 2.00 bpw set and IQ2_XS the 2.25 bpw set, 4 shards each, plus ONE Engram table (110.6 GB, the same file for both sets:
+# downloaded once into the shared folder) and the DSpark drafter and the 5.8 MB metadata shard of Unsloth's set that pack_gguf --engram-meta reads
+# (both pinned in DSV41_FILES above). This is the set a normal PC runs (64 GB RAM + a 12 GB card; docs/DEEPSEEK41.md); the UD-Q8_K_XL entry stays
+# for the big machines.
+DSV41_ISTA_REPO = "ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF"
+DSV41_ISTA_ENGRAM = "DeepSeek-V4.1-Flash-Engram-IQ4_NL.gguf"
+DSV41_ISTA_SHARD = "DeepSeek-V4.1-Flash-1017-GSQ-RCO-{quant}-{i:05d}-of-00004.gguf"
DSV41_DSPARK = "dspark-DeepSeek-V4.1-Flash-1017-Q8_0.gguf"
DSV41_ISTA_FILES = {
+ "DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XS-00001-of-00004.gguf": (39535776992, "4c65307992ab586c527ca1f6f64062aa96a33b23c02b7fed103f45f75e854798"),
+ "DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XS-00002-of-00004.gguf": (39494617536, "933461a974fcccee741bf2f9798201e72ab4aa3f8a952be56f849896ca6ea30e"),
+ "DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XS-00003-of-00004.gguf": (39423838496, "bbf91fa346a85ebaf32dd4fd5c456b534d780d381e54f46ac90783a481373240"),
+ "DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XS-00004-of-00004.gguf": (36771591232, "e5b2f8781fab688cd4bc3bc7a69651b5dbb90ed94965657c6bead8bb4774f3cf"),
+ "DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00001-of-00004.gguf": (39075714624, "d46efb67ccffd8fb5d7f18b423896381b9e06d8491e161c370ddf3b05b07e050"),
+ "DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00002-of-00004.gguf": (39388449504, "cec1b2c3da429dc768b8bb9bc4dc4403d81f74d34db660a6a34ee3e555bf2c70"),
+ "DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00003-of-00004.gguf": (39246891616, "9d9c2140fcb134a0b9a74dedd0e5167e68077af6d08ce7f856bce112f0a35dd3"),
+ "DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00004-of-00004.gguf": (20527837312, "9b790a2297ed3cdede379fc984f4a52bb95f3b23724c797af11d1fa6a69db059"),
+ "DeepSeek-V4.1-Flash-Engram-IQ4_NL.gguf": (110595290816, "eb19f30d368d393784e1fc54fdfaffc427ebb2110bf3762fdf5a8fbb9a2d740a"),
+}
+# The files the first release of these sets was downloaded as (helcig/tmp-oct26-2, 2026-10-23: model-2.00bpw-0000N-of-00004.gguf, model-2.25bpw-...,
+# engram-IQ4_NL.gguf, usually in a 2.00bpw/ or 2.25bpw/ folder): testers and our boxes have them. They are the same weights: shards 2-4 are the
+# same bytes as today's, and shard 1 and the Engram differ only in their GGUF header (metadata). setup accepts them from "files already here"
+# (size and SHA-256 checked against these pins) and never downloads them. Not in REPO_PINS: nothing is fetched under these names.
+DSV41_LEGACY_FILES = {
"model-2.00bpw-00001-of-00004.gguf": (39075713728, "78ab9efc46a26647c3e94b322c99cd7e2af261dd28b38f2712e128562db1ef47"),
"model-2.00bpw-00002-of-00004.gguf": (39388449504, "cec1b2c3da429dc768b8bb9bc4dc4403d81f74d34db660a6a34ee3e555bf2c70"),
"model-2.00bpw-00003-of-00004.gguf": (39246891616, "9d9c2140fcb134a0b9a74dedd0e5167e68077af6d08ce7f856bce112f0a35dd3"),
@@ -697,13 +713,19 @@ DSV41_ISTA_FILES = {
"model-2.25bpw-00004-of-00004.gguf": (36771591232, "e5b2f8781fab688cd4bc3bc7a69651b5dbb90ed94965657c6bead8bb4774f3cf"),
"engram-IQ4_NL.gguf": (110595290720, "1f17d8ae37e9df3bb45f7ef1b4fcd1706d7b759af0413c117bf8e92aed459281"),
}
+# today's name -> the name it had in the first release
+DSV41_LEGACY_NAMES = {
+ **{DSV41_ISTA_SHARD.format(quant="IQ2_XXS", i=i): "model-2.00bpw-%05d-of-00004.gguf" % i for i in range(1, 5)},
+ **{DSV41_ISTA_SHARD.format(quant="IQ2_XS", i=i): "model-2.25bpw-%05d-of-00004.gguf" % i for i in range(1, 5)},
+ DSV41_ISTA_ENGRAM: "engram-IQ4_NL.gguf",
+}
# the files every official set shares (name -> where it comes from); they live in models/deepseek-v4.1-flash-shared, once
DSV41_SHARED = {
- DSV41_ISTA_ENGRAM: hf(DSV41_ISTA_REPO) + "2.00bpw/",
+ DSV41_ISTA_ENGRAM: hf(DSV41_ISTA_REPO),
DSV41_SHARD.format(i=1): hf(DSV41_REPO) + "UD-Q8_K_XL/", # only its metadata is read (pack_gguf --engram-meta)
DSV41_DSPARK: hf(DSV41_REPO),
}
-DSV41_ALL_FILES = {**DSV41_FILES, **DSV41_ISTA_FILES}
+DSV41_ALL_FILES = {**DSV41_FILES, **DSV41_ISTA_FILES, **DSV41_LEGACY_FILES}
DSV41_PROFILES = ROOT / "data" / "dsv41" # the shipped hot-expert profiles (profile-2.00.bin, profile-2.25.bin, + .sha256)
DSV41_CONTEXT = 8192 # the context the engine was checked at (4096 and 8192; 64K is not tried yet)
MULTI_FILE_FAMILIES = {
@@ -725,15 +747,15 @@ MULTI_FILE_FAMILIES["deepseek41"]["models"] = {
"about": "the smaller official version: 138 GB of model + 111 GB n-gram table + 8 GB drafter (257 GB of disk). Runs on a PC with "
"a 12 GB card and 40-64 GB of RAM at about 5-7 tokens a second; quality about 16% above the original model's "
"perplexity",
- "ram_gb": 40, "recipe": "ista", "subdir": "2.00bpw", "profile": "profile-2.00.bin", "hf": hf(DSV41_ISTA_REPO) + "2.00bpw/",
- "shards": [DSV41_ISTA_SHARD.format(bpw="2.00bpw", i=i) for i in range(1, 5)],
+ "ram_gb": 40, "recipe": "ista", "subdir": None, "legacy_subdir": "2.00bpw", "profile": "profile-2.00.bin", "hf": hf(DSV41_ISTA_REPO),
+ "shards": [DSV41_ISTA_SHARD.format(quant="IQ2_XXS", i=i) for i in range(1, 5)],
"shared": [DSV41_ISTA_ENGRAM, DSV41_SHARD.format(i=1), DSV41_DSPARK]},
"IST-DAS-2.25bpw": {
"label": "IST-DAS 2.25 bpw",
"about": "the better official version: 155 GB of model + 111 GB n-gram table + 8 GB drafter (274 GB of disk, 17 GB more than "
"2.00). About the same speed; quality about 8% above the original model's perplexity (code and agent work 2.5-3.5%)",
- "ram_gb": 44, "recipe": "ista", "subdir": "2.25bpw", "profile": "profile-2.25.bin", "hf": hf(DSV41_ISTA_REPO) + "2.25bpw/",
- "shards": [DSV41_ISTA_SHARD.format(bpw="2.25bpw", i=i) for i in range(1, 5)],
+ "ram_gb": 44, "recipe": "ista", "subdir": None, "legacy_subdir": "2.25bpw", "profile": "profile-2.25.bin", "hf": hf(DSV41_ISTA_REPO),
+ "shards": [DSV41_ISTA_SHARD.format(quant="IQ2_XS", i=i) for i in range(1, 5)],
"shared": [DSV41_ISTA_ENGRAM, DSV41_SHARD.format(i=1), DSV41_DSPARK]},
"UD-Q8_K_XL": {
"label": "Unsloth UD-Q8_K_XL",
@@ -754,7 +776,7 @@ REPO_PINS = {
"unsloth/Qwen3.6-35B-A3B-MTP-GGUF": QWEN36_FILES,
"agentionai/Qwen3.8-Flash-Next-Gyro-GGUF": GYRO_FILES,
"unsloth/DeepSeek-V4.1-Flash-1017-GGUF": DSV41_FILES,
- "helcig/tmp-oct26-2": DSV41_ISTA_FILES,
+ "ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF": DSV41_ISTA_FILES,
}
# EXPERIMENTAL, off by default (setup asks): a control vector shipped with the repository, see its README
ESP_VECTOR = ROOT / "data" / "experimental-speed-projection" / "Qwen3.8-Flash-Next-experimental-speed-projection.gguf"
@@ -7768,11 +7790,27 @@ def dsv41_folder(a, fam: dict, tag: str, model: str = "UD-Q8_K_XL") -> Path:
if a.gguf_dir:
folder = Path(a.gguf_dir)
d = fam["models"][model]
- sub = d.get("subdir") or model
- return folder / sub if (folder / sub).is_dir() and not (folder / d["shards"][0]).exists() else folder
+ names = [n for s in d["shards"] for n in (s, DSV41_LEGACY_NAMES.get(s)) if n]
+ if any((folder / n).exists() for n in names):
+ return folder
+ for sub in (d.get("subdir"), d.get("legacy_subdir"), model): # the set's own folder inside (the first release used 2.00bpw/, 2.25bpw/)
+ if sub and (folder / sub).is_dir():
+ return folder / sub
+ return folder
return Path(a.models_dir) / tag
+def dsv41_pick(folder: Path, name: str) -> Path:
+ """`name` in `folder`, or the file it was called in the first release (DSV41_LEGACY_NAMES; the same weights) when only that one is there;
+ else `name` (where it will be downloaded to)."""
+ if (folder / name).exists():
+ return folder / name
+ old = DSV41_LEGACY_NAMES.get(name)
+ if old and (folder / old).exists():
+ return folder / old
+ return folder / name
+
+
def dsv41_shared_dir(a, fam: dict) -> Path:
"""Where the files the official sets share (the Engram, the DSpark drafter, the metadata shard) are downloaded to: one folder, so
installing the second set does not download the 111 GB Engram again."""
@@ -7783,7 +7821,7 @@ def dsv41_find_shared(a, name: str, folder: Path, shared: Path) -> Path:
"""The file `name` (an Engram, the drafter, ...) the person already has: in the shared folder, beside the shards, or (--gguf-dir) in the
folder above them or one folder level down from either; else where it will be downloaded to (the shared folder). The size must be the
pinned one, so a half-copied file is not taken for the file."""
- size = DSV41_ALL_FILES[name][0]
+ names = [n for n in (name, DSV41_LEGACY_NAMES.get(name)) if n] # today's name, and the first release's (the same weights)
roots = [shared, folder]
if a.gguf_dir:
base = Path(a.gguf_dir)
@@ -7798,12 +7836,13 @@ def dsv41_find_shared(a, name: str, folder: Path, shared: Path) -> Path:
except OSError:
continue
for k in kids:
- f = k / name
- try:
- if f.is_file() and f.stat().st_size == size:
- return f
- except OSError:
- pass
+ for n in names:
+ f = k / n
+ try:
+ if f.is_file() and f.stat().st_size == DSV41_ALL_FILES[n][0]:
+ return f
+ except OSError:
+ pass
return shared / name
@@ -7883,14 +7922,14 @@ def install_dsv41(a, family, gpu, hip, amd, cuda_tk, multi, port, ram, adopted,
model = names[int(ask("Which version?", [str(i) for i in range(1, len(names) + 1)], "1", a.yes)) - 1]
ok(f"size: {model}")
md = fam["models"][model]
- say(" Its GGUF files: " + ("IST-DAS (huggingface.co/helcig/tmp-oct26-2), with the DSpark drafter and a small metadata file from Unsloth"
+ say(" Its GGUF files: " + ("IST-DAS / ISTA-DASLab (huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF), with the DSpark drafter and a small metadata file from Unsloth"
if md.get("recipe") == "ista" else "Unsloth (huggingface.co/unsloth/DeepSeek-V4.1-Flash-1017-GGUF)"))
ista = md.get("recipe") == "ista"
tag = fam["tag"] + model
cards = dsv41_cards(gpu, all_cards, bool(a.gpu is not None or a.gpus))
ctx = a.context or DSV41_CONTEXT
folder = dsv41_folder(a, fam, tag, model)
- shards = [folder / n for n in md["shards"]]
+ shards = [dsv41_pick(folder, n) for n in md["shards"]]
shared_dir = dsv41_shared_dir(a, fam)
shared = [dsv41_find_shared(a, n, folder, shared_dir) for n in md.get("shared", [])]
own_files = {f: bool(a.gguf_dir) or f.parent != shared_dir for f in shared} # a file found outside setup's own folder is the person's
@@ -7899,7 +7938,7 @@ def install_dsv41(a, family, gpu, hip, amd, cuda_tk, multi, port, ram, adopted,
p = f.with_name(f.name + ".part")
return p.stat().st_size if p.is_file() else 0
- left = sum(0 if f.exists() and (done(f) or a.gguf_dir or own_files.get(f)) else fam["sha256"][f.name][0] - partial(f)
+ left = sum(0 if f.exists() and (done(f) or a.gguf_dir or own_files.get(f) or f.name in DSV41_LEGACY_FILES) else fam["sha256"][f.name][0] - partial(f)
for f in [*shards, *shared]) / 1e9
probe = folder if folder.exists() else (folder.parent if folder.parent.exists() else Path(a.models_dir))
free = free_gb(probe)
@@ -7968,9 +8007,11 @@ def install_dsv41(a, family, gpu, hip, amd, cuda_tk, multi, port, ram, adopted,
else:
if a.gguf_dir and not f.exists():
fail(f"{f} is missing", f"put {f.name} in {folder}, or drop --gguf-dir to download it")
- if not a.gguf_dir:
+ if not a.gguf_dir and f.name not in DSV41_LEGACY_FILES: # (a first-release file is checked, never fetched under its old name)
download((md.get("hf") or fam["hf"]) + f.name, f)
- verify_sha256(f, *fam["sha256"][f.name], own=bool(a.gguf_dir))
+ if f.name in DSV41_LEGACY_FILES:
+ ok(f"{f.name}: a file of the first release of this set (the same weights); it is used as it is")
+ verify_sha256(f, *fam["sha256"][f.name], own=bool(a.gguf_dir) or f.name in DSV41_LEGACY_FILES)
for f in shared: # the Engram, the drafter, the metadata shard: once for every official set
if f.exists() and (done(f) or own_files[f]):
ok(f"{f.name} already here ({f.parent})")
@@ -7991,7 +8032,7 @@ def install_dsv41(a, family, gpu, hip, amd, cuda_tk, multi, port, ram, adopted,
have_pack = False
if not have_pack:
if ista:
- run(dsv41_pack_cmd(folder, md["shards"], shared[0], shared[1], shared[2], pack))
+ run(dsv41_pack_cmd(folder, [f.name for f in shards], shared[0], shared[1], shared[2], pack))
else:
run([sys.executable, str(ROOT / "tools" / "dsv41" / "pack_gguf.py"), "--model", str(folder), "--out", str(pack)])
if ista:
diff --git a/strata_control/dsv41_fit.py b/strata_control/dsv41_fit.py
index 96b841123..b5dc68572 100644
--- a/strata_control/dsv41_fit.py
+++ b/strata_control/dsv41_fit.py
@@ -48,20 +48,20 @@ MEASURED = ("4 x Radeon AI PRO R9700 (32 GB) with 251 GB of RAM: prompts of 3,00
"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
+# The two official IST-DAS sets (ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF, 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 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_ENGRAM_BYTES = 110_595_290_816
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",
+ "IST-DAS-2.00bpw": {"title": "IST-DAS 2.00 bpw", "set_bytes": 138_238_893_056, "experts_gb": 132.07, "profile": "profile-2.00.bin",
"ppl": "16%", "ram_ok_gb": 40.0, "ram_great_gb": 160.0, "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",
+ "IST-DAS-2.25bpw": {"title": "IST-DAS 2.25 bpw", "set_bytes": 155_225_824_256, "experts_gb": 149.06, "profile": "profile-2.25.bin",
"ppl": "8%", "ram_ok_gb": 44.0, "ram_great_gb": 176.0, "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"},
diff --git a/tools/test_setup_dsv41.py b/tools/test_setup_dsv41.py
index add4bf8c1..6d951379f 100644
--- a/tools/test_setup_dsv41.py
+++ b/tools/test_setup_dsv41.py
@@ -322,32 +322,35 @@ RTX5070 = {"index": 0, "name": "NVIDIA GeForce RTX 5070", "vram_gb": 12.0, "vend
class OfficialSets(unittest.TestCase):
- """The two official IST-DAS sets (helcig/tmp-oct26-2): the catalog, the pins, the pack recipe, the fit words."""
+ """The two official IST-DAS sets (ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF): the catalog, the pins, the pack recipe, the fit words."""
def test_the_pinned_files(self):
fam = setup.MULTI_FILE_FAMILIES["deepseek41"]
self.assertEqual(list(fam["models"]), ["IST-DAS-2.00bpw", "IST-DAS-2.25bpw", "UD-Q8_K_XL"]) # the official sets first
- self.assertEqual(setup.HF_REVISIONS["helcig/tmp-oct26-2"], "851709dcf0d97f091a4bdce758b2e4b0b468dd26")
- self.assertIs(setup.REPO_PINS["helcig/tmp-oct26-2"], setup.DSV41_ISTA_FILES)
+ repo = "ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF"
+ self.assertEqual(setup.HF_REVISIONS[repo], "0d8f49fb15df73789a9c237a1b8b7b088423dbc0")
+ self.assertIs(setup.REPO_PINS[repo], setup.DSV41_ISTA_FILES)
+ self.assertNotIn("helcig/tmp-oct26-2", setup.HF_REVISIONS) # the first release's repository is only remembered as local names
+ self.assertEqual(setup.DSV41_ISTA_REPO, repo)
self.assertEqual(len(setup.DSV41_ISTA_FILES), 9) # 2 x 4 shards + the Engram
for n, (size, sha) in setup.DSV41_ISTA_FILES.items():
self.assertRegex(sha, r"^[0-9a-f]{64}$", n)
- eng = setup.DSV41_ISTA_FILES["engram-IQ4_NL.gguf"]
- self.assertEqual(eng, (110_595_290_720, "1f17d8ae37e9df3bb45f7ef1b4fcd1706d7b759af0413c117bf8e92aed459281"))
+ eng = setup.DSV41_ISTA_FILES["DeepSeek-V4.1-Flash-Engram-IQ4_NL.gguf"]
+ self.assertEqual(eng, (110_595_290_816, "eb19f30d368d393784e1fc54fdfaffc427ebb2110bf3762fdf5a8fbb9a2d740a"))
self.assertEqual(setup.DSV41_ALL_FILES["dspark-DeepSeek-V4.1-Flash-1017-Q8_0.gguf"],
(7_976_914_240, "6ba6ac1b331c540691a209e7fc97dec2a24e295acb1e755c84c7d24a4373927a"))
sums = {m: sum(setup.DSV41_ALL_FILES[n][0] for n in fam["models"][m]["shards"]) for m in list(fam["models"])[:2]}
- self.assertEqual(sums, {"IST-DAS-2.00bpw": 138_238_892_160, "IST-DAS-2.25bpw": 155_225_823_328})
+ self.assertEqual(sums, {"IST-DAS-2.00bpw": 138238893056, "IST-DAS-2.25bpw": 155225824256})
self.assertEqual(sums["IST-DAS-2.00bpw"], dsv41_fit.ISTA_SETS["IST-DAS-2.00bpw"]["set_bytes"])
self.assertEqual(sums["IST-DAS-2.25bpw"], dsv41_fit.ISTA_SETS["IST-DAS-2.25bpw"]["set_bytes"])
- self.assertEqual(setup.DSV41_ISTA_FILES["engram-IQ4_NL.gguf"][0], dsv41_fit.ISTA_ENGRAM_BYTES)
+ self.assertEqual(setup.DSV41_ISTA_FILES["DeepSeek-V4.1-Flash-Engram-IQ4_NL.gguf"][0], dsv41_fit.ISTA_ENGRAM_BYTES)
for m in ("IST-DAS-2.00bpw", "IST-DAS-2.25bpw"): # both sets name the same shared files: the Engram is downloaded once
self.assertEqual(fam["models"][m]["shared"], [setup.DSV41_ISTA_ENGRAM, setup.DSV41_SHARD.format(i=1), setup.DSV41_DSPARK])
for n in fam["models"][m]["shared"]:
self.assertIn(n, setup.DSV41_SHARED)
self.assertIn(n, setup.DSV41_ALL_FILES)
- self.assertRegex(setup.DSV41_SHARED[setup.DSV41_ISTA_ENGRAM], r"helcig/tmp-oct26-2/resolve/851709dc[0-9a-f]{32}/2\.00bpw/$")
- self.assertRegex(fam["models"]["IST-DAS-2.25bpw"]["hf"], r"helcig/tmp-oct26-2/resolve/851709dc[0-9a-f]{32}/2\.25bpw/$")
+ self.assertRegex(setup.DSV41_SHARED[setup.DSV41_ISTA_ENGRAM], r"ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/resolve/0d8f49fb[0-9a-f]{32}/$")
+ self.assertRegex(fam["models"]["IST-DAS-2.25bpw"]["hf"], r"ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/resolve/0d8f49fb[0-9a-f]{32}/$")
def test_shipped_profiles_match_their_hashes(self):
import hashlib
@@ -359,7 +362,7 @@ class OfficialSets(unittest.TestCase):
def test_the_pack_recipe_is_the_plain_one(self):
folder, pack = Path("/m/set"), Path("/d/pack")
- names = [f"model-2.00bpw-{i:05d}-of-00004.gguf" for i in range(1, 5)]
+ names = [f"DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-{i:05d}-of-00004.gguf" for i in range(1, 5)]
cmd = setup.dsv41_pack_cmd(folder, names, Path("/s/engram-IQ4_NL.gguf"), Path("/s/meta.gguf"), Path("/s/dspark.gguf"), pack)
self.assertTrue(cmd[1].endswith("pack_gguf.py"))
self.assertEqual(cmd[2:], ["--model", str(folder), "--shards", *names, "--engram", str(Path("/s/engram-IQ4_NL.gguf")),
@@ -418,6 +421,53 @@ class SharedFiles(unittest.TestCase):
self.assertEqual(got, f)
+class LegacyFiles(unittest.TestCase):
+ """Files of the first release (model-2.00bpw-0000N-of-00004.gguf, engram-IQ4_NL.gguf, often in a 2.00bpw/ folder) are the same weights: accepted."""
+
+ def test_the_old_pins_are_the_old_files_and_the_unchanged_shards_match_today_s(self):
+ for i in (2, 3, 4): # same bytes: same size and SHA-256 under both names
+ new = setup.DSV41_ISTA_SHARD.format(quant="IQ2_XXS", i=i)
+ self.assertEqual(setup.DSV41_ALL_FILES[new], setup.DSV41_LEGACY_FILES["model-2.00bpw-%05d-of-00004.gguf" % i])
+ new = setup.DSV41_ISTA_SHARD.format(quant="IQ2_XS", i=i)
+ self.assertEqual(setup.DSV41_ALL_FILES[new], setup.DSV41_LEGACY_FILES["model-2.25bpw-%05d-of-00004.gguf" % i])
+ self.assertNotEqual(setup.DSV41_ALL_FILES[setup.DSV41_ISTA_SHARD.format(quant="IQ2_XXS", i=1)],
+ setup.DSV41_LEGACY_FILES["model-2.00bpw-00001-of-00004.gguf"]) # shard 1 and the Engram differ in their header only
+ self.assertNotIn("model-2.00bpw-00001-of-00004.gguf", setup.DSV41_ISTA_FILES) # (never downloaded under the old names)
+ self.assertNotIn("model-2.00bpw-00001-of-00004.gguf", setup.REPO_PINS["ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF"])
+
+ def test_pick_prefers_today_s_name_then_the_old_one(self):
+ with tempfile.TemporaryDirectory() as d:
+ folder = Path(d)
+ new = setup.DSV41_ISTA_SHARD.format(quant="IQ2_XXS", i=2)
+ self.assertEqual(setup.dsv41_pick(folder, new), folder / new) # nothing there: where it will be downloaded
+ (folder / "model-2.00bpw-00002-of-00004.gguf").write_bytes(b"x")
+ self.assertEqual(setup.dsv41_pick(folder, new).name, "model-2.00bpw-00002-of-00004.gguf")
+ (folder / new).write_bytes(b"y")
+ self.assertEqual(setup.dsv41_pick(folder, new).name, new)
+
+ def test_the_folder_of_the_first_release_is_found_from_the_parent(self):
+ with tempfile.TemporaryDirectory() as d:
+ sub = Path(d) / "2.00bpw"
+ sub.mkdir()
+ (sub / "model-2.00bpw-00001-of-00004.gguf").write_bytes(b"x")
+ fam = setup.MULTI_FILE_FAMILIES["deepseek41"]
+ a = argparse.Namespace(gguf_dir=d, models_dir=d)
+ self.assertEqual(setup.dsv41_folder(a, fam, "t", "IST-DAS-2.00bpw"), sub)
+ a2 = argparse.Namespace(gguf_dir=str(sub), models_dir=d)
+ self.assertEqual(setup.dsv41_folder(a2, fam, "t", "IST-DAS-2.00bpw"), sub)
+
+ def test_the_old_engram_name_is_found(self):
+ with tempfile.TemporaryDirectory() as d:
+ sub = Path(d) / "2.00bpw"
+ sub.mkdir()
+ old = sub / "engram-IQ4_NL.gguf"
+ old.write_bytes(b"e")
+ a = argparse.Namespace(gguf_dir=d)
+ with mock.patch.dict(setup.DSV41_ALL_FILES, {"engram-IQ4_NL.gguf": (1, "0" * 64)}):
+ got = setup.dsv41_find_shared(a, setup.DSV41_ISTA_ENGRAM, sub, Path(d) / "shared")
+ self.assertEqual(got, old)
+
+
class InstallOfficial(Install):
"""install_dsv41 for an official set, the engine and the tools mocked."""
@@ -438,10 +488,11 @@ class InstallOfficial(Install):
get = [c for c in self.calls if c[0] == "get"]
folder = self.root / "models" / "deepseek-v4.1-flash-IST-DAS-2.00bpw"
shared = self.root / "models" / "deepseek-v4.1-flash-shared"
- self.assertEqual([c[2] for c in get][:4], [folder / f"model-2.00bpw-{i:05d}-of-00004.gguf" for i in range(1, 5)])
+ self.assertEqual([c[2] for c in get][:4], [folder / f"DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-{i:05d}-of-00004.gguf" for i in range(1, 5)])
self.assertEqual([c[2] for c in get][4:], [shared / setup.DSV41_ISTA_ENGRAM, shared / setup.DSV41_SHARD.format(i=1),
shared / setup.DSV41_DSPARK])
- self.assertTrue(get[0][1].endswith("/2.00bpw/model-2.00bpw-00001-of-00004.gguf"))
+ self.assertTrue(get[0][1].endswith("/DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00001-of-00004.gguf"))
+ self.assertNotIn("/2.00bpw/", get[0][1]) # the repository is flat
self.assertEqual(sum(1 for c in self.calls if c[0] == "sha"), 7)
pack = [c for c in self.calls if c[0] == "run"][-1][1]
self.assertIn("--engram-meta", pack)
@@ -459,7 +510,7 @@ class InstallOfficial(Install):
rc, out = self.install_ista(model="IST-DAS-2.25bpw")
self.assertEqual(rc, 0, out)
got = [c[2].name for c in self.calls if c[0] == "get"]
- self.assertEqual(got, [f"model-2.25bpw-{i:05d}-of-00004.gguf" for i in range(1, 5)]) # only the four shards
+ self.assertEqual(got, [f"DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XS-{i:05d}-of-00004.gguf" for i in range(1, 5)]) # only the four shards
def test_pack_info_profile_smoke_and_the_launch_flags(self):
rc, out = self.install_ista()