joemaddalone commited on
Commit
e6a8d26
·
verified ·
1 Parent(s): d548d67

GGUF q8_0 export with --no-mtp/--disable-vision fallbacks

Browse files
Files changed (1) hide show
  1. scripts/gguf_export.py +13 -3
scripts/gguf_export.py CHANGED
@@ -11,6 +11,7 @@ def sh(cmd, **kw):
11
 
12
  # deps for llama.cpp's converter
13
  sh("pip install -q --no-cache-dir gguf numpy 'torch<3' transformers sentencepiece protobuf || true")
 
14
  if not os.path.isdir(f"{WORK}/llama.cpp"):
15
  sh(f"git clone --depth 1 https://github.com/ggml-org/llama.cpp {WORK}/llama.cpp")
16
 
@@ -19,9 +20,18 @@ local = snapshot_download(REPO)
19
  print("model at", local, flush=True)
20
 
21
  out = f"{WORK}/ffmpeg-nl2cmd-0.8B-q8_0.gguf"
22
- r = sh(f"python {WORK}/llama.cpp/convert_hf_to_gguf.py {local} --outfile {out} --outtype q8_0")
23
- if r.returncode != 0:
24
- print("CONVERT_FAILED", flush=True)
 
 
 
 
 
 
 
 
 
25
  sys.exit(1)
26
  print("gguf size:", os.path.getsize(out), flush=True)
27
 
 
11
 
12
  # deps for llama.cpp's converter
13
  sh("pip install -q --no-cache-dir gguf numpy 'torch<3' transformers sentencepiece protobuf || true")
14
+ sh("apt-get update -qq && apt-get install -y -qq git >/dev/null 2>&1 || true")
15
  if not os.path.isdir(f"{WORK}/llama.cpp"):
16
  sh(f"git clone --depth 1 https://github.com/ggml-org/llama.cpp {WORK}/llama.cpp")
17
 
 
20
  print("model at", local, flush=True)
21
 
22
  out = f"{WORK}/ffmpeg-nl2cmd-0.8B-q8_0.gguf"
23
+ import sys
24
+ ok = False
25
+ for extra in ([], ["--no-mtp"], ["--no-mtp", "--disable-vision"]):
26
+ r = sh(f"python {WORK}/llama.cpp/convert_hf_to_gguf.py {local} --outfile {out} --outtype q8_0 " +
27
+ " ".join(extra))
28
+ if r.returncode == 0 and os.path.exists(out):
29
+ print("converted with extra args:", extra, flush=True)
30
+ ok = True
31
+ break
32
+ print("conversion failed with extra args:", extra, flush=True)
33
+ if not ok:
34
+ print("CONVERT_FAILED (all fallbacks)", flush=True)
35
  sys.exit(1)
36
  print("gguf size:", os.path.getsize(out), flush=True)
37