streamft 0.5.0__tar.gz → 0.6.1__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {streamft-0.5.0 → streamft-0.6.1}/PKG-INFO +49 -3
- {streamft-0.5.0 → streamft-0.6.1}/README.md +48 -2
- {streamft-0.5.0 → streamft-0.6.1}/pyproject.toml +1 -1
- {streamft-0.5.0 → streamft-0.6.1}/streamft/__main__.py +6 -0
- streamft-0.6.1/streamft/adapter.py +150 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/lora.py +28 -9
- {streamft-0.5.0 → streamft-0.6.1}/streamft/serve.py +11 -3
- {streamft-0.5.0 → streamft-0.6.1}/streamft.egg-info/PKG-INFO +49 -3
- {streamft-0.5.0 → streamft-0.6.1}/streamft.egg-info/SOURCES.txt +2 -0
- streamft-0.6.1/tests/test_adapter_quant.py +84 -0
- {streamft-0.5.0 → streamft-0.6.1}/LICENSE +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/setup.cfg +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/__init__.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/autotune.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/cli.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/data.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/disk_quant_store.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/doctor.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/easy.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/evals.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/fit.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/merge.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/metrics.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/stream_quant_lora.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft/stream_train.py +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft.egg-info/dependency_links.txt +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft.egg-info/entry_points.txt +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft.egg-info/requires.txt +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/streamft.egg-info/top_level.txt +0 -0
- {streamft-0.5.0 → streamft-0.6.1}/tests/test_streamft_smoke.py +0 -0
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: streamft
|
|
3
|
-
Version: 0.
|
|
3
|
+
Version: 0.6.1
|
|
4
4
|
Summary: Fine-tune models that do not fit: stream NF4 weights from disk, one transformer block at a time.
|
|
5
5
|
Author-email: Gokhan Cava <byjameson@users.noreply.github.com>
|
|
6
6
|
License: Apache-2.0
|
|
@@ -150,8 +150,9 @@ streamft train --model-dir checkpoints/qwen7b_nf4 --sft --steps 1000 \
|
|
|
150
150
|
- On the reference Ampere card fp16 is slightly faster than bf16 at 7B (1.79 vs 1.91 s/step,
|
|
151
151
|
same 5.36 GB peak); on Pascal it is the only option.
|
|
152
152
|
- On a **4 GB** card use `--blocks-on-gpu 1`, `--seq 512` and `--vram-hard-gb 3.4`.
|
|
153
|
-
Those rows are not measured yet — if you run one,
|
|
154
|
-
|
|
153
|
+
Those rows are not measured yet — if you run one, open a
|
|
154
|
+
[hardware report](.github/ISSUE_TEMPLATE/hardware-report.md) and it can become a measured
|
|
155
|
+
entry in `streamft fit`.
|
|
155
156
|
- `streamft fit` prints this hint automatically when it sees a pre-Ampere GPU.
|
|
156
157
|
|
|
157
158
|
### Tokenizer cost and the diffusion question
|
|
@@ -191,6 +192,41 @@ tools can point at it. This is **one machine computing, the rest as clients** ?
|
|
|
191
192
|
peer-to-peer sharding and not volunteer compute. Tested end to end: `/health` and a
|
|
192
193
|
16-token completion (1.9 s on the reference GPU).
|
|
193
194
|
|
|
195
|
+
### Share one base, ship 5 MB adapters
|
|
196
|
+
|
|
197
|
+
A LoRA adapter is the per-behaviour artifact. Pack it to 4 bits and serve it on top of a
|
|
198
|
+
shared base:
|
|
199
|
+
|
|
200
|
+
```bash
|
|
201
|
+
streamft adapter --adapter checkpoints/my_adapter.pt --bits 4 --out checkpoints/my_adapter_q4.pt
|
|
202
|
+
streamft serve --model checkpoints/qwen3b_bnb4bit --adapter checkpoints/my_adapter_q4.pt
|
|
203
|
+
```
|
|
204
|
+
|
|
205
|
+
Measured on the reference machine (Qwen2.5-3B, 60-step Dolly LoRA):
|
|
206
|
+
|
|
207
|
+
| storage | size | bits / base-param |
|
|
208
|
+
|---|---:|---:|
|
|
209
|
+
| fp32 | 29.6 MB | 0.0766 |
|
|
210
|
+
| fp16 | 14.8 MB | 0.0382 |
|
|
211
|
+
| 8-bit | 8.56 MB | 0.0222 |
|
|
212
|
+
| **4-bit** | **4.88 MB** | **0.0126** |
|
|
213
|
+
|
|
214
|
+
870 held-out Dolly rows, masked SFT loss: base 2.3752, original 2.3403, 8-bit 2.3401
|
|
215
|
+
(p = 0.81, no measurable loss), 4-bit 2.3420 (+0.0018, p = 0.001, 5% of the gain). A
|
|
216
|
+
second task, SST-2 (2,000 train / 400 test), goes from **71.25%** base to **93.75%**, and
|
|
217
|
+
both packed adapters stay at **93.75%** (masked loss 0.7940/0.7942 vs 0.7930; p = 0.50 /
|
|
218
|
+
0.41). A third task, AG News (4-way), goes from 2.50% with the bare prompt (not followed)
|
|
219
|
+
to **62.25%**; 8-bit and 4-bit stay at **62.00%**. The adapter loads in 0.115 s. The LoRA
|
|
220
|
+
path is ~26% slower than the base on this card (8.75 vs 11.79 tok/s; a bf16 LoRA path
|
|
221
|
+
reaches 9.13); merge it into a bf16 base for zero per-token overhead, or accept the LoRA
|
|
222
|
+
path when storage/switching matters more than throughput.
|
|
223
|
+
Rank reduction is not viable: SVD to rank 8 loses 35.8% of the update.
|
|
224
|
+
|
|
225
|
+
**A prompt can be the better rung.** On the same base, a 441-byte 4-shot SST-2 prompt
|
|
226
|
+
scores 94.0% (adapter 93.75%) and a 242-byte 4-shot AG News prompt scores 61.75% (adapter
|
|
227
|
+
62.25%). If a few examples express the task, use the prompt; the adapter is for behaviour
|
|
228
|
+
a prompt cannot express, or when prompt tokens/context cost more than the storage.
|
|
229
|
+
|
|
194
230
|
### Measured on
|
|
195
231
|
|
|
196
232
|
Every speed number here comes from **one machine**: RTX 3050 Laptop 6 GB, Windows 11,
|
|
@@ -207,6 +243,16 @@ derives the architecture fingerprint and prints the snippets for `fit.py`, `fitc
|
|
|
207
243
|
and the Space. Anything not in that table prints **"not measured"** in `streamft fit` ?
|
|
208
244
|
that is the point.
|
|
209
245
|
|
|
246
|
+
### Why 0.1-bit compression is not a universal target
|
|
247
|
+
|
|
248
|
+
A measured note on quantization vs distillation: post-hoc quantization dies below 3 bits;
|
|
249
|
+
a 340 KB student matches a 1.5B teacher on a narrow keyword task at 0.0019
|
|
250
|
+
bit/teacher-param; the same 4.4M student is within 1.3 points on DBPedia-14 but loses
|
|
251
|
+
22.5 on SST-2 and 47.9 on MNLI; open-ended generation dies at 3 bits. A cheap 45 KB TF-IDF
|
|
252
|
+
student is a **lower bound** on the 4.4M student's normalized skill (6 tasks, R² = 0.89),
|
|
253
|
+
not a predictor: QQP (C 0.21, S 0.49) breaks the near-identity. Details, caveats and
|
|
254
|
+
reproduction: [`docs/BEHAVIORAL_BITS.md`](docs/BEHAVIORAL_BITS.md).
|
|
255
|
+
|
|
210
256
|
## Verification
|
|
211
257
|
|
|
212
258
|
`python scripts/verify_gradients.py --model-dir checkpoints/qwen3b_bnb4bit --seq 128`
|
|
@@ -120,8 +120,9 @@ streamft train --model-dir checkpoints/qwen7b_nf4 --sft --steps 1000 \
|
|
|
120
120
|
- On the reference Ampere card fp16 is slightly faster than bf16 at 7B (1.79 vs 1.91 s/step,
|
|
121
121
|
same 5.36 GB peak); on Pascal it is the only option.
|
|
122
122
|
- On a **4 GB** card use `--blocks-on-gpu 1`, `--seq 512` and `--vram-hard-gb 3.4`.
|
|
123
|
-
Those rows are not measured yet — if you run one,
|
|
124
|
-
|
|
123
|
+
Those rows are not measured yet — if you run one, open a
|
|
124
|
+
[hardware report](.github/ISSUE_TEMPLATE/hardware-report.md) and it can become a measured
|
|
125
|
+
entry in `streamft fit`.
|
|
125
126
|
- `streamft fit` prints this hint automatically when it sees a pre-Ampere GPU.
|
|
126
127
|
|
|
127
128
|
### Tokenizer cost and the diffusion question
|
|
@@ -161,6 +162,41 @@ tools can point at it. This is **one machine computing, the rest as clients** ?
|
|
|
161
162
|
peer-to-peer sharding and not volunteer compute. Tested end to end: `/health` and a
|
|
162
163
|
16-token completion (1.9 s on the reference GPU).
|
|
163
164
|
|
|
165
|
+
### Share one base, ship 5 MB adapters
|
|
166
|
+
|
|
167
|
+
A LoRA adapter is the per-behaviour artifact. Pack it to 4 bits and serve it on top of a
|
|
168
|
+
shared base:
|
|
169
|
+
|
|
170
|
+
```bash
|
|
171
|
+
streamft adapter --adapter checkpoints/my_adapter.pt --bits 4 --out checkpoints/my_adapter_q4.pt
|
|
172
|
+
streamft serve --model checkpoints/qwen3b_bnb4bit --adapter checkpoints/my_adapter_q4.pt
|
|
173
|
+
```
|
|
174
|
+
|
|
175
|
+
Measured on the reference machine (Qwen2.5-3B, 60-step Dolly LoRA):
|
|
176
|
+
|
|
177
|
+
| storage | size | bits / base-param |
|
|
178
|
+
|---|---:|---:|
|
|
179
|
+
| fp32 | 29.6 MB | 0.0766 |
|
|
180
|
+
| fp16 | 14.8 MB | 0.0382 |
|
|
181
|
+
| 8-bit | 8.56 MB | 0.0222 |
|
|
182
|
+
| **4-bit** | **4.88 MB** | **0.0126** |
|
|
183
|
+
|
|
184
|
+
870 held-out Dolly rows, masked SFT loss: base 2.3752, original 2.3403, 8-bit 2.3401
|
|
185
|
+
(p = 0.81, no measurable loss), 4-bit 2.3420 (+0.0018, p = 0.001, 5% of the gain). A
|
|
186
|
+
second task, SST-2 (2,000 train / 400 test), goes from **71.25%** base to **93.75%**, and
|
|
187
|
+
both packed adapters stay at **93.75%** (masked loss 0.7940/0.7942 vs 0.7930; p = 0.50 /
|
|
188
|
+
0.41). A third task, AG News (4-way), goes from 2.50% with the bare prompt (not followed)
|
|
189
|
+
to **62.25%**; 8-bit and 4-bit stay at **62.00%**. The adapter loads in 0.115 s. The LoRA
|
|
190
|
+
path is ~26% slower than the base on this card (8.75 vs 11.79 tok/s; a bf16 LoRA path
|
|
191
|
+
reaches 9.13); merge it into a bf16 base for zero per-token overhead, or accept the LoRA
|
|
192
|
+
path when storage/switching matters more than throughput.
|
|
193
|
+
Rank reduction is not viable: SVD to rank 8 loses 35.8% of the update.
|
|
194
|
+
|
|
195
|
+
**A prompt can be the better rung.** On the same base, a 441-byte 4-shot SST-2 prompt
|
|
196
|
+
scores 94.0% (adapter 93.75%) and a 242-byte 4-shot AG News prompt scores 61.75% (adapter
|
|
197
|
+
62.25%). If a few examples express the task, use the prompt; the adapter is for behaviour
|
|
198
|
+
a prompt cannot express, or when prompt tokens/context cost more than the storage.
|
|
199
|
+
|
|
164
200
|
### Measured on
|
|
165
201
|
|
|
166
202
|
Every speed number here comes from **one machine**: RTX 3050 Laptop 6 GB, Windows 11,
|
|
@@ -177,6 +213,16 @@ derives the architecture fingerprint and prints the snippets for `fit.py`, `fitc
|
|
|
177
213
|
and the Space. Anything not in that table prints **"not measured"** in `streamft fit` ?
|
|
178
214
|
that is the point.
|
|
179
215
|
|
|
216
|
+
### Why 0.1-bit compression is not a universal target
|
|
217
|
+
|
|
218
|
+
A measured note on quantization vs distillation: post-hoc quantization dies below 3 bits;
|
|
219
|
+
a 340 KB student matches a 1.5B teacher on a narrow keyword task at 0.0019
|
|
220
|
+
bit/teacher-param; the same 4.4M student is within 1.3 points on DBPedia-14 but loses
|
|
221
|
+
22.5 on SST-2 and 47.9 on MNLI; open-ended generation dies at 3 bits. A cheap 45 KB TF-IDF
|
|
222
|
+
student is a **lower bound** on the 4.4M student's normalized skill (6 tasks, R² = 0.89),
|
|
223
|
+
not a predictor: QQP (C 0.21, S 0.49) breaks the near-identity. Details, caveats and
|
|
224
|
+
reproduction: [`docs/BEHAVIORAL_BITS.md`](docs/BEHAVIORAL_BITS.md).
|
|
225
|
+
|
|
180
226
|
## Verification
|
|
181
227
|
|
|
182
228
|
`python scripts/verify_gradients.py --model-dir checkpoints/qwen3b_bnb4bit --seq 128`
|
|
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
|
|
|
4
4
|
|
|
5
5
|
[project]
|
|
6
6
|
name = "streamft"
|
|
7
|
-
version = "0.
|
|
7
|
+
version = "0.6.1"
|
|
8
8
|
description = "Fine-tune models that do not fit: stream NF4 weights from disk, one transformer block at a time."
|
|
9
9
|
readme = "README.md"
|
|
10
10
|
requires-python = ">=3.10"
|
|
@@ -8,6 +8,7 @@
|
|
|
8
8
|
streamft synth generate training pairs locally with a small model
|
|
9
9
|
streamft eval run a small local prompt set (repetition/refusal/keywords)
|
|
10
10
|
streamft merge fold a saved adapter into the base model (for serving / GGUF)
|
|
11
|
+
streamft adapter pack a saved adapter to 8/4-bit (shared-base deployment)
|
|
11
12
|
streamft serve host a merged model on the LAN (chat page + /v1/chat/completions)
|
|
12
13
|
streamft doctor environment report (torch, CUDA, bitsandbytes, VRAM)
|
|
13
14
|
"""
|
|
@@ -25,6 +26,7 @@ USAGE = """streamft - fine-tune models that do not fit
|
|
|
25
26
|
streamft synth --model <small-model> --n 50 --out data/synth.jsonl
|
|
26
27
|
streamft eval --model <model> [--prompts mine.jsonl]
|
|
27
28
|
streamft merge --model <base> --adapter <dir> --out merged/<name>
|
|
29
|
+
streamft adapter --adapter <pt> --bits 4 --out <pt>
|
|
28
30
|
streamft serve --model merged/<name> --port 8000
|
|
29
31
|
streamft doctor
|
|
30
32
|
|
|
@@ -57,6 +59,10 @@ def main() -> int:
|
|
|
57
59
|
from streamft.merge import main as merge_main
|
|
58
60
|
return merge_main(rest)
|
|
59
61
|
|
|
62
|
+
if cmd == "adapter":
|
|
63
|
+
from streamft.adapter import main as adapter_main
|
|
64
|
+
return adapter_main(rest)
|
|
65
|
+
|
|
60
66
|
if cmd == "eval":
|
|
61
67
|
from streamft.evals import main as eval_main
|
|
62
68
|
return eval_main(rest)
|
|
@@ -0,0 +1,150 @@
|
|
|
1
|
+
"""Quantized LoRA adapter storage (8-bit / 4-bit, group-wise).
|
|
2
|
+
|
|
3
|
+
The adapter is the per-behaviour artifact in `docs/BEHAVIORAL_BITS.md`: a shared base
|
|
4
|
+
plus a few MB of adapter. This module stores the A/B tensors as packed integers with
|
|
5
|
+
per-group scales, and dequantizes them at load time. It is deliberately separate from the
|
|
6
|
+
trainer: the adapter format is a deployment format, not a training format.
|
|
7
|
+
|
|
8
|
+
streamft adapter --adapter checkpoints/my_adapter.pt --bits 4 --out q4.pt
|
|
9
|
+
|
|
10
|
+
The 4-bit file is smaller than fp32 but carries per-group scales; the real file size is
|
|
11
|
+
what counts, not the packed-code size alone.
|
|
12
|
+
"""
|
|
13
|
+
from __future__ import annotations
|
|
14
|
+
|
|
15
|
+
import argparse
|
|
16
|
+
import json
|
|
17
|
+
from pathlib import Path
|
|
18
|
+
|
|
19
|
+
import torch
|
|
20
|
+
|
|
21
|
+
FORMAT = "streamft-adapter-q1"
|
|
22
|
+
|
|
23
|
+
|
|
24
|
+
def _quantize(w: torch.Tensor, bits: int, group: int):
|
|
25
|
+
flat = w.reshape(-1).float()
|
|
26
|
+
n = flat.numel()
|
|
27
|
+
pad = (-n) % group
|
|
28
|
+
if pad:
|
|
29
|
+
flat = torch.cat([flat, torch.zeros(pad)])
|
|
30
|
+
g = flat.view(-1, group)
|
|
31
|
+
lo = g.amin(1, keepdim=True)
|
|
32
|
+
hi = g.amax(1, keepdim=True)
|
|
33
|
+
levels = 2 ** bits - 1
|
|
34
|
+
scale = (hi - lo).clamp_min(1e-12) / levels
|
|
35
|
+
code = ((g - lo) / scale).round().clamp(0, levels).to(torch.uint8)
|
|
36
|
+
if bits == 8:
|
|
37
|
+
packed = code.reshape(-1)
|
|
38
|
+
elif bits == 4:
|
|
39
|
+
c = code.reshape(-1, 2)
|
|
40
|
+
packed = (c[:, 0] | (c[:, 1] << 4)).to(torch.uint8)
|
|
41
|
+
elif bits == 2:
|
|
42
|
+
c = code.reshape(-1, 4)
|
|
43
|
+
packed = (c[:, 0] | (c[:, 1] << 2) | (c[:, 2] << 4) | (c[:, 3] << 6)).to(torch.uint8)
|
|
44
|
+
else:
|
|
45
|
+
raise ValueError("bits must be 2, 4 or 8")
|
|
46
|
+
return {"packed": packed, "lo": lo.squeeze(1), "scale": scale.squeeze(1),
|
|
47
|
+
"n": n, "shape": list(w.shape)}
|
|
48
|
+
|
|
49
|
+
|
|
50
|
+
def _unpack(q: dict, bits: int) -> torch.Tensor:
|
|
51
|
+
p = q["packed"]
|
|
52
|
+
if bits == 8:
|
|
53
|
+
code = p
|
|
54
|
+
elif bits == 4:
|
|
55
|
+
code = torch.stack([p & 0xF, (p >> 4) & 0xF], dim=1).reshape(-1)
|
|
56
|
+
elif bits == 2:
|
|
57
|
+
code = torch.stack([p & 0x3, (p >> 2) & 0x3, (p >> 4) & 0x3, (p >> 6) & 0x3],
|
|
58
|
+
dim=1).reshape(-1)
|
|
59
|
+
else:
|
|
60
|
+
raise ValueError("bits must be 2, 4 or 8")
|
|
61
|
+
return code
|
|
62
|
+
|
|
63
|
+
|
|
64
|
+
def _dequantize(q: dict, bits: int) -> torch.Tensor:
|
|
65
|
+
# _unpack returns the padded code vector (groups * group); slice after dequantizing.
|
|
66
|
+
code = _unpack(q, bits).float()
|
|
67
|
+
group = int(q.get("group", q["scale"].numel()))
|
|
68
|
+
scale = q["scale"].float()
|
|
69
|
+
lo = q["lo"].float()
|
|
70
|
+
flat = code.view(-1, group)
|
|
71
|
+
out = flat * scale[:, None] + lo[:, None]
|
|
72
|
+
return out.reshape(-1)[: q["n"]].reshape(q["shape"])
|
|
73
|
+
|
|
74
|
+
|
|
75
|
+
def quantize_adapter(state: dict, bits: int, group: int = 64, meta: dict | None = None) -> dict:
|
|
76
|
+
tensors = {}
|
|
77
|
+
for name, w in state.items():
|
|
78
|
+
if name.endswith(".A") or name.endswith(".B"):
|
|
79
|
+
q = _quantize(w, bits, group)
|
|
80
|
+
q["group"] = group
|
|
81
|
+
tensors[name] = q
|
|
82
|
+
return {"format": FORMAT, "bits": bits, "group": group,
|
|
83
|
+
"meta": meta or {}, "tensors": tensors}
|
|
84
|
+
|
|
85
|
+
|
|
86
|
+
def dequantize_adapter(obj: dict) -> dict:
|
|
87
|
+
return {name: _dequantize(q, obj["bits"]) for name, q in obj["tensors"].items()}
|
|
88
|
+
|
|
89
|
+
|
|
90
|
+
def save_quantized(state: dict, path: str | Path, bits: int, group: int = 64,
|
|
91
|
+
meta: dict | None = None) -> dict:
|
|
92
|
+
obj = quantize_adapter(state, bits, group, meta)
|
|
93
|
+
torch.save(obj, str(path))
|
|
94
|
+
return obj
|
|
95
|
+
|
|
96
|
+
|
|
97
|
+
def load_adapter(path: str | Path):
|
|
98
|
+
"""Load a plain .pt adapter ({"lora": ...}) or a quantized one; return (state, meta)."""
|
|
99
|
+
obj = torch.load(str(path), map_location="cpu", weights_only=True)
|
|
100
|
+
if isinstance(obj, dict) and obj.get("format") == FORMAT:
|
|
101
|
+
return dequantize_adapter(obj), obj.get("meta", {})
|
|
102
|
+
if isinstance(obj, dict) and "lora" in obj:
|
|
103
|
+
return obj["lora"], obj.get("meta", {})
|
|
104
|
+
if isinstance(obj, dict):
|
|
105
|
+
return obj, {}
|
|
106
|
+
raise ValueError(f"not an adapter: {path}")
|
|
107
|
+
|
|
108
|
+
|
|
109
|
+
def tensor_error(a: torch.Tensor, b: torch.Tensor) -> float:
|
|
110
|
+
return float((a.float() - b.float()).norm() / a.float().norm().clamp_min(1e-12))
|
|
111
|
+
|
|
112
|
+
|
|
113
|
+
def main(argv=None) -> int:
|
|
114
|
+
ap = argparse.ArgumentParser(prog="streamft adapter")
|
|
115
|
+
ap.add_argument("--adapter", required=True,
|
|
116
|
+
help="adapter .pt, or a save dir containing adapter.pt")
|
|
117
|
+
ap.add_argument("--bits", type=int, choices=[8, 4, 2], default=4,
|
|
118
|
+
help="8=safe, 4=default, 2=not behaviourally validated")
|
|
119
|
+
ap.add_argument("--group", type=int, default=64)
|
|
120
|
+
ap.add_argument("--out", required=True)
|
|
121
|
+
args = ap.parse_args(argv)
|
|
122
|
+
src = Path(args.adapter)
|
|
123
|
+
if src.is_dir():
|
|
124
|
+
src = src / "adapter.pt"
|
|
125
|
+
state, meta = load_adapter(src)
|
|
126
|
+
if not state:
|
|
127
|
+
raise SystemExit(f"no tensors in adapter: {src}")
|
|
128
|
+
lora_keys = {k for k in state if k.endswith(".A") or k.endswith(".B")}
|
|
129
|
+
if not lora_keys:
|
|
130
|
+
raise SystemExit(f"no LoRA A/B tensors in {src}; nothing to pack")
|
|
131
|
+
skipped = sorted(set(state) - lora_keys)
|
|
132
|
+
out = Path(args.out)
|
|
133
|
+
obj = save_quantized(state, out, args.bits, args.group, meta)
|
|
134
|
+
back = dequantize_adapter(obj)
|
|
135
|
+
errs = [tensor_error(state[k], back[k]) for k in back]
|
|
136
|
+
params = sum(v.numel() for v in state.values())
|
|
137
|
+
print(json.dumps({
|
|
138
|
+
"adapter": str(src), "out": str(out), "bits": args.bits, "group": args.group,
|
|
139
|
+
"params": params,
|
|
140
|
+
"src_bytes": src.stat().st_size, "out_bytes": out.stat().st_size,
|
|
141
|
+
"compression": round(src.stat().st_size / max(out.stat().st_size, 1), 2),
|
|
142
|
+
"max_relative_error": round(max(errs) if errs else 0.0, 4),
|
|
143
|
+
"tensors": len(back),
|
|
144
|
+
"skipped_tensors": skipped,
|
|
145
|
+
}, indent=1))
|
|
146
|
+
return 0
|
|
147
|
+
|
|
148
|
+
|
|
149
|
+
if __name__ == "__main__":
|
|
150
|
+
raise SystemExit(main())
|
|
@@ -26,21 +26,40 @@ class LoRALinear(nn.Module):
|
|
|
26
26
|
self.r = r
|
|
27
27
|
self.scale = alpha / r
|
|
28
28
|
w = base.weight
|
|
29
|
-
|
|
30
|
-
|
|
31
|
-
self.
|
|
29
|
+
# Use in_features/out_features, not weight.shape: bitsandbytes 4-bit layers
|
|
30
|
+
# store a packed weight whose shape is not (out, in).
|
|
31
|
+
self.A = nn.Parameter(torch.randn(r, base.in_features, device=w.device,
|
|
32
|
+
dtype=torch.float32) / math.sqrt(base.in_features))
|
|
33
|
+
self.B = nn.Parameter(torch.zeros(base.out_features, r, device=w.device, dtype=torch.float32))
|
|
32
34
|
self.dropout = nn.Dropout(dropout) if dropout > 0 else nn.Identity()
|
|
35
|
+
self.fast_dtype = None # inference-only bf16/fp16 path; training stays fp32
|
|
33
36
|
|
|
34
37
|
def forward(self, x):
|
|
35
38
|
y = self.base(x)
|
|
36
|
-
|
|
37
|
-
|
|
38
|
-
|
|
39
|
-
|
|
40
|
-
|
|
41
|
-
|
|
39
|
+
if self.fast_dtype is not None:
|
|
40
|
+
h = x.to(self.fast_dtype)
|
|
41
|
+
A = self.A.to(device=h.device, dtype=self.fast_dtype)
|
|
42
|
+
B = self.B.to(device=h.device, dtype=self.fast_dtype)
|
|
43
|
+
else:
|
|
44
|
+
h = self.dropout(x).float()
|
|
45
|
+
# A/B live on CPU (grads are accumulated there and the optimizer steps
|
|
46
|
+
# there); hop them to the activation's device for the matmul. The low-rank
|
|
47
|
+
# path is tiny, so this costs a negligible amount of PCIe traffic.
|
|
48
|
+
A = self.A if self.A.device == h.device else self.A.to(h.device)
|
|
49
|
+
B = self.B if self.B.device == h.device else self.B.to(h.device)
|
|
42
50
|
return y + (self.scale * (h @ A.t() @ B.t())).to(y.dtype)
|
|
43
51
|
|
|
52
|
+
@torch.no_grad()
|
|
53
|
+
def set_fast(self, dtype=torch.bfloat16):
|
|
54
|
+
"""Inference-only: run the low-rank path in `dtype` instead of fp32.
|
|
55
|
+
|
|
56
|
+
The adapter is trained in fp32 (small values would vanish in fp16); this is a
|
|
57
|
+
deployment switch. The weights are rounded once, at load time.
|
|
58
|
+
"""
|
|
59
|
+
self.A.data = self.A.data.to(dtype)
|
|
60
|
+
self.B.data = self.B.data.to(dtype)
|
|
61
|
+
self.fast_dtype = dtype
|
|
62
|
+
|
|
44
63
|
@torch.no_grad()
|
|
45
64
|
def merge(self):
|
|
46
65
|
self.base.weight.data += (self.scale * (self.B @ self.A)).to(self.base.weight.dtype)
|
|
@@ -108,7 +108,8 @@ def main(argv=None) -> int:
|
|
|
108
108
|
|
|
109
109
|
ap = argparse.ArgumentParser(prog="streamft serve")
|
|
110
110
|
ap.add_argument("--model", required=True, help="merged model dir or HF id")
|
|
111
|
-
ap.add_argument("--adapter", default="",
|
|
111
|
+
ap.add_argument("--adapter", default="",
|
|
112
|
+
help="optional adapter .pt, or a save dir containing adapter.pt")
|
|
112
113
|
ap.add_argument("--host", default="0.0.0.0")
|
|
113
114
|
ap.add_argument("--port", type=int, default=8000)
|
|
114
115
|
ap.add_argument("--device", default="cuda" if torch.cuda.is_available() else "cpu")
|
|
@@ -121,14 +122,21 @@ def main(argv=None) -> int:
|
|
|
121
122
|
if args.adapter:
|
|
122
123
|
from pathlib import Path
|
|
123
124
|
|
|
124
|
-
from .
|
|
125
|
+
from .adapter import load_adapter
|
|
126
|
+
from .lora import LoRALinear, apply_lora, set_lora_state_dict
|
|
125
127
|
|
|
126
128
|
p = Path(args.adapter)
|
|
127
129
|
p = p / "adapter.pt" if p.is_dir() else p
|
|
128
|
-
state =
|
|
130
|
+
state, _meta = load_adapter(p)
|
|
129
131
|
r = int(next(v for k, v in state.items() if k.endswith(".A")).shape[0])
|
|
130
132
|
apply_lora(model, r_attn=r, r_mlp=0)
|
|
131
133
|
set_lora_state_dict(model, state)
|
|
134
|
+
if dtype != torch.float32:
|
|
135
|
+
# Inference-only bf16 low-rank path: ~4% faster than fp32 on the reference
|
|
136
|
+
# card and it matches the base dtype; training still uses fp32.
|
|
137
|
+
for m in model.modules():
|
|
138
|
+
if isinstance(m, LoRALinear):
|
|
139
|
+
m.set_fast(dtype)
|
|
132
140
|
print(f"[serve] adapter {p.name} (r={r})", flush=True)
|
|
133
141
|
|
|
134
142
|
_state.update(model=model, tok=tok, device=args.device, model_id=args.model)
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: streamft
|
|
3
|
-
Version: 0.
|
|
3
|
+
Version: 0.6.1
|
|
4
4
|
Summary: Fine-tune models that do not fit: stream NF4 weights from disk, one transformer block at a time.
|
|
5
5
|
Author-email: Gokhan Cava <byjameson@users.noreply.github.com>
|
|
6
6
|
License: Apache-2.0
|
|
@@ -150,8 +150,9 @@ streamft train --model-dir checkpoints/qwen7b_nf4 --sft --steps 1000 \
|
|
|
150
150
|
- On the reference Ampere card fp16 is slightly faster than bf16 at 7B (1.79 vs 1.91 s/step,
|
|
151
151
|
same 5.36 GB peak); on Pascal it is the only option.
|
|
152
152
|
- On a **4 GB** card use `--blocks-on-gpu 1`, `--seq 512` and `--vram-hard-gb 3.4`.
|
|
153
|
-
Those rows are not measured yet — if you run one,
|
|
154
|
-
|
|
153
|
+
Those rows are not measured yet — if you run one, open a
|
|
154
|
+
[hardware report](.github/ISSUE_TEMPLATE/hardware-report.md) and it can become a measured
|
|
155
|
+
entry in `streamft fit`.
|
|
155
156
|
- `streamft fit` prints this hint automatically when it sees a pre-Ampere GPU.
|
|
156
157
|
|
|
157
158
|
### Tokenizer cost and the diffusion question
|
|
@@ -191,6 +192,41 @@ tools can point at it. This is **one machine computing, the rest as clients** ?
|
|
|
191
192
|
peer-to-peer sharding and not volunteer compute. Tested end to end: `/health` and a
|
|
192
193
|
16-token completion (1.9 s on the reference GPU).
|
|
193
194
|
|
|
195
|
+
### Share one base, ship 5 MB adapters
|
|
196
|
+
|
|
197
|
+
A LoRA adapter is the per-behaviour artifact. Pack it to 4 bits and serve it on top of a
|
|
198
|
+
shared base:
|
|
199
|
+
|
|
200
|
+
```bash
|
|
201
|
+
streamft adapter --adapter checkpoints/my_adapter.pt --bits 4 --out checkpoints/my_adapter_q4.pt
|
|
202
|
+
streamft serve --model checkpoints/qwen3b_bnb4bit --adapter checkpoints/my_adapter_q4.pt
|
|
203
|
+
```
|
|
204
|
+
|
|
205
|
+
Measured on the reference machine (Qwen2.5-3B, 60-step Dolly LoRA):
|
|
206
|
+
|
|
207
|
+
| storage | size | bits / base-param |
|
|
208
|
+
|---|---:|---:|
|
|
209
|
+
| fp32 | 29.6 MB | 0.0766 |
|
|
210
|
+
| fp16 | 14.8 MB | 0.0382 |
|
|
211
|
+
| 8-bit | 8.56 MB | 0.0222 |
|
|
212
|
+
| **4-bit** | **4.88 MB** | **0.0126** |
|
|
213
|
+
|
|
214
|
+
870 held-out Dolly rows, masked SFT loss: base 2.3752, original 2.3403, 8-bit 2.3401
|
|
215
|
+
(p = 0.81, no measurable loss), 4-bit 2.3420 (+0.0018, p = 0.001, 5% of the gain). A
|
|
216
|
+
second task, SST-2 (2,000 train / 400 test), goes from **71.25%** base to **93.75%**, and
|
|
217
|
+
both packed adapters stay at **93.75%** (masked loss 0.7940/0.7942 vs 0.7930; p = 0.50 /
|
|
218
|
+
0.41). A third task, AG News (4-way), goes from 2.50% with the bare prompt (not followed)
|
|
219
|
+
to **62.25%**; 8-bit and 4-bit stay at **62.00%**. The adapter loads in 0.115 s. The LoRA
|
|
220
|
+
path is ~26% slower than the base on this card (8.75 vs 11.79 tok/s; a bf16 LoRA path
|
|
221
|
+
reaches 9.13); merge it into a bf16 base for zero per-token overhead, or accept the LoRA
|
|
222
|
+
path when storage/switching matters more than throughput.
|
|
223
|
+
Rank reduction is not viable: SVD to rank 8 loses 35.8% of the update.
|
|
224
|
+
|
|
225
|
+
**A prompt can be the better rung.** On the same base, a 441-byte 4-shot SST-2 prompt
|
|
226
|
+
scores 94.0% (adapter 93.75%) and a 242-byte 4-shot AG News prompt scores 61.75% (adapter
|
|
227
|
+
62.25%). If a few examples express the task, use the prompt; the adapter is for behaviour
|
|
228
|
+
a prompt cannot express, or when prompt tokens/context cost more than the storage.
|
|
229
|
+
|
|
194
230
|
### Measured on
|
|
195
231
|
|
|
196
232
|
Every speed number here comes from **one machine**: RTX 3050 Laptop 6 GB, Windows 11,
|
|
@@ -207,6 +243,16 @@ derives the architecture fingerprint and prints the snippets for `fit.py`, `fitc
|
|
|
207
243
|
and the Space. Anything not in that table prints **"not measured"** in `streamft fit` ?
|
|
208
244
|
that is the point.
|
|
209
245
|
|
|
246
|
+
### Why 0.1-bit compression is not a universal target
|
|
247
|
+
|
|
248
|
+
A measured note on quantization vs distillation: post-hoc quantization dies below 3 bits;
|
|
249
|
+
a 340 KB student matches a 1.5B teacher on a narrow keyword task at 0.0019
|
|
250
|
+
bit/teacher-param; the same 4.4M student is within 1.3 points on DBPedia-14 but loses
|
|
251
|
+
22.5 on SST-2 and 47.9 on MNLI; open-ended generation dies at 3 bits. A cheap 45 KB TF-IDF
|
|
252
|
+
student is a **lower bound** on the 4.4M student's normalized skill (6 tasks, R² = 0.89),
|
|
253
|
+
not a predictor: QQP (C 0.21, S 0.49) breaks the near-identity. Details, caveats and
|
|
254
|
+
reproduction: [`docs/BEHAVIORAL_BITS.md`](docs/BEHAVIORAL_BITS.md).
|
|
255
|
+
|
|
210
256
|
## Verification
|
|
211
257
|
|
|
212
258
|
`python scripts/verify_gradients.py --model-dir checkpoints/qwen3b_bnb4bit --seq 128`
|
|
@@ -3,6 +3,7 @@ README.md
|
|
|
3
3
|
pyproject.toml
|
|
4
4
|
streamft/__init__.py
|
|
5
5
|
streamft/__main__.py
|
|
6
|
+
streamft/adapter.py
|
|
6
7
|
streamft/autotune.py
|
|
7
8
|
streamft/cli.py
|
|
8
9
|
streamft/data.py
|
|
@@ -23,4 +24,5 @@ streamft.egg-info/dependency_links.txt
|
|
|
23
24
|
streamft.egg-info/entry_points.txt
|
|
24
25
|
streamft.egg-info/requires.txt
|
|
25
26
|
streamft.egg-info/top_level.txt
|
|
27
|
+
tests/test_adapter_quant.py
|
|
26
28
|
tests/test_streamft_smoke.py
|
|
@@ -0,0 +1,84 @@
|
|
|
1
|
+
import torch
|
|
2
|
+
from torch import nn
|
|
3
|
+
|
|
4
|
+
from streamft.adapter import load_adapter, save_quantized, tensor_error
|
|
5
|
+
from streamft.lora import LoRALinear
|
|
6
|
+
|
|
7
|
+
|
|
8
|
+
def _state():
|
|
9
|
+
torch.manual_seed(0)
|
|
10
|
+
return {
|
|
11
|
+
"model.layers.0.self_attn.q_proj.A": torch.randn(8, 2048) * 0.02,
|
|
12
|
+
"model.layers.0.self_attn.q_proj.B": torch.randn(2048, 8) * 0.02,
|
|
13
|
+
}
|
|
14
|
+
|
|
15
|
+
|
|
16
|
+
def test_roundtrip_8bit(tmp_path):
|
|
17
|
+
state = _state()
|
|
18
|
+
path = tmp_path / "a8.pt"
|
|
19
|
+
save_quantized(state, path, 8, group=64)
|
|
20
|
+
back, meta = load_adapter(path)
|
|
21
|
+
assert set(back) == set(state)
|
|
22
|
+
for k in state:
|
|
23
|
+
assert back[k].shape == state[k].shape
|
|
24
|
+
assert tensor_error(state[k], back[k]) < 0.02
|
|
25
|
+
assert path.stat().st_size < sum(v.numel() for v in state.values()) * 4
|
|
26
|
+
|
|
27
|
+
|
|
28
|
+
def test_roundtrip_4bit(tmp_path):
|
|
29
|
+
state = _state()
|
|
30
|
+
path = tmp_path / "a4.pt"
|
|
31
|
+
save_quantized(state, path, 4, group=64)
|
|
32
|
+
back, _ = load_adapter(path)
|
|
33
|
+
for k in state:
|
|
34
|
+
assert back[k].shape == state[k].shape
|
|
35
|
+
assert tensor_error(state[k], back[k]) < 0.15
|
|
36
|
+
assert path.stat().st_size < sum(v.numel() for v in state.values()) * 2
|
|
37
|
+
|
|
38
|
+
|
|
39
|
+
def test_plain_adapter_load(tmp_path):
|
|
40
|
+
state = _state()
|
|
41
|
+
path = tmp_path / "plain.pt"
|
|
42
|
+
torch.save({"lora": state, "meta": {"tag": "x"}}, path)
|
|
43
|
+
back, meta = load_adapter(path)
|
|
44
|
+
assert set(back) == set(state)
|
|
45
|
+
assert meta["tag"] == "x"
|
|
46
|
+
|
|
47
|
+
|
|
48
|
+
def test_lora_linear_shapes():
|
|
49
|
+
m = LoRALinear(nn.Linear(8, 4), r=2)
|
|
50
|
+
assert m.A.shape == (2, 8)
|
|
51
|
+
assert m.B.shape == (4, 2)
|
|
52
|
+
assert m(torch.randn(3, 8)).shape == (3, 4)
|
|
53
|
+
|
|
54
|
+
|
|
55
|
+
def test_cli_accepts_save_dir(tmp_path):
|
|
56
|
+
from streamft.adapter import main
|
|
57
|
+
|
|
58
|
+
save_dir = tmp_path / "run"
|
|
59
|
+
save_dir.mkdir()
|
|
60
|
+
torch.save(_state(), save_dir / "adapter.pt")
|
|
61
|
+
out = tmp_path / "adapter_q8.pt"
|
|
62
|
+
assert main(["--adapter", str(save_dir), "--bits", "8", "--out", str(out)]) == 0
|
|
63
|
+
assert out.exists()
|
|
64
|
+
back, _ = load_adapter(out)
|
|
65
|
+
assert set(back) == set(_state())
|
|
66
|
+
|
|
67
|
+
|
|
68
|
+
def test_cli_rejects_non_lora_state(tmp_path):
|
|
69
|
+
import pytest
|
|
70
|
+
|
|
71
|
+
from streamft.adapter import main
|
|
72
|
+
|
|
73
|
+
path = tmp_path / "plain.pt"
|
|
74
|
+
torch.save({"weight": torch.zeros(4, 4)}, path)
|
|
75
|
+
with pytest.raises(SystemExit, match="no LoRA"):
|
|
76
|
+
main(["--adapter", str(path), "--bits", "8", "--out", str(tmp_path / "q.pt")])
|
|
77
|
+
|
|
78
|
+
|
|
79
|
+
def test_lora_fast_path():
|
|
80
|
+
m = LoRALinear(nn.Linear(8, 4).to(torch.bfloat16), r=2)
|
|
81
|
+
m.set_fast(torch.bfloat16)
|
|
82
|
+
out = m(torch.randn(3, 8, dtype=torch.bfloat16))
|
|
83
|
+
assert out.shape == (3, 4)
|
|
84
|
+
assert out.dtype == torch.bfloat16
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|