Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[wip] AMD Benchmarks #1383

Open
wants to merge 9 commits into
base: main
Choose a base branch
from
+137 −69
Open

[wip] AMD Benchmarks #1383

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
a609805
wip
jcaip Dec 5, 2024
aec9414
update requirements
jcaip Dec 5, 2024
24ec597
new
jcaip Dec 5, 2024
edad9f9
updated
jcaip Dec 5, 2024
421709d
fix typo
jcaip Dec 5, 2024
e374ac0
wip
jcaip Dec 5, 2024
f66b270
fix
jcaip Dec 5, 2024
e12bf45
unpin mkl libs
jcaip Dec 5, 2024
6a30ecd
remove mkl libs
jcaip Dec 5, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
49 changes: 49 additions & 0 deletions benchmark_requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,49 @@
attrs==24.2.0
black==24.8.0
blobfile==3.0.0
certifi==2024.8.30
cffi==1.17.1
charset-normalizer==3.4.0
click==8.1.7
expecttest==0.2.1
filelock==3.16.1
fsspec==2024.10.0
huggingface-hub==0.26.2
hypothesis==6.115.6
idna==3.10
Jinja2==3.1.4
lxml==5.3.0
markdown-it-py==2.2.0
MarkupSafe==2.1.5
mdurl==0.1.0
mpmath==1.3.0
mypy==1.11.2
mypy-extensions==1.0.0
networkx==3.4.2
numpy==2.0.1
packaging==24.1
pathspec==0.10.3
pillow==11.0.0
pip==24.3.1
platformdirs==3.10.0
protobuf==4.25.3
psutil==5.9.0
pycparser==2.21
pycryptodomex==3.21.0
Pygments==2.15.1
PyYAML==6.0.2
regex==2024.11.6
requests==2.32.3
rich==13.7.1
safetensors==0.4.5
setuptools==75.1.0
six==1.16.0
sortedcontainers==2.4.0
sympy==1.13.1
tiktoken==0.8.0
tokenizers==0.20.3
tqdm==4.67.1
transformers==4.46.3
typing_extensions==4.11.0
urllib3==2.2.3
wheel==0.44.0
5 changes: 5 additions & 0 deletions pytorch_requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
--index-url https://download.pytorch.org/whl/nightly/rocm6.2
torch==2.6.0.dev20241122+rocm6.2
pytorch-triton-rocm==3.1.0+cf34004b8a
torchaudio==2.5.0.dev20241125+rocm6.2
torchvision==0.20.0.dev20241125+rocm6.2
4 changes: 2 additions & 2 deletions scripts/convert_hf_checkpoint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,8 +86,8 @@ def permute(w, n_head):
state_dict = torch.load(str(file), map_location="cpu", mmap=True, weights_only=True)
merged_result.update(state_dict)

if config.tie_word_embeddings:
merged_result["lm_head.weight"] = merged_result["model.embed_tokens.weight"].clone()
# if config.tie_word_embeddings:
# merged_result["lm_head.weight"] = merged_result["model.embed_tokens.weight"].clone()

final_result = {}
for key, value in merged_result.items():
Expand Down
12 changes: 6 additions & 6 deletions scripts/prepare.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,8 +1,8 @@
python scripts/download.py --repo_id meta-llama/Llama-2-7b-chat-hf
python scripts/download.py --repo_id meta-llama/Meta-Llama-3-8B
#python scripts/download.py --repo_id meta-llama/Llama-2-7b-chat-hf
#python scripts/download.py --repo_id meta-llama/Meta-Llama-3-8B
python scripts/download.py --repo_id meta-llama/Meta-Llama-3.1-8B
python scripts/download.py --repo_id meta-llama/Llama-3.2-3B
python scripts/convert_hf_checkpoint.py --checkpoint_dir checkpoints/meta-llama/Llama-2-7b-chat-hf
python scripts/convert_hf_checkpoint.py --checkpoint_dir checkpoints/meta-llama/Meta-Llama-3-8B
#python scripts/download.py --repo_id meta-llama/Llama-3.2-3B
#python scripts/convert_hf_checkpoint.py --checkpoint_dir checkpoints/meta-llama/Llama-2-7b-chat-hf
#python scripts/convert_hf_checkpoint.py --checkpoint_dir checkpoints/meta-llama/Meta-Llama-3-8B
python scripts/convert_hf_checkpoint.py --checkpoint_dir checkpoints/meta-llama/Meta-Llama-3.1-8B
python scripts/convert_hf_checkpoint.py --checkpoint_dir checkpoints/meta-llama/Llama-3.2-3B
#python scripts/convert_hf_checkpoint.py --checkpoint_dir checkpoints/meta-llama/Llama-3.2-3B
14 changes: 14 additions & 0 deletions torchao/_models/llama/benchmark_results.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,3 +50,17 @@ OTHER BENCHMARKS
20240910010056, tok/s= 47.85, mem/s= 213.24 GB/s, peak_mem=11.85 GB, model_size= 4.46 GB quant: uintx-4-64, mod: Meta-Llama-3-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization uintx-4-64 --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --top_k 200 --temperature 0.8
20240910010647, tok/s= 34.83, mem/s= 261.42 GB/s, peak_mem=14.99 GB, model_size= 7.51 GB quant: uintx-2-8, mod: Meta-Llama-3-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization uintx-2-8 --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --top_k 200 --temperature 0.8
20240910110958, tok/s=223.95, mem/s= 682.88 GB/s, peak_mem= 5.59 GB, model_size= 3.05 GB quant: sparse-marlin, mod: Meta-Llama-3-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.float16, device: cuda repro: python generate.py --quantization sparse-marlin --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3-8B/model.pth --device cuda --precision torch.float16 --compile --num_samples 5 --max_new_tokens 200 --top_k 200 --temperature 0.8

20241125122729, tok/s=160.77, mem/s=2413.14 GB/s, peak_mem=16.60 GB, model_size=15.01 GB quant: None, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125123143, tok/s=213.99, mem/s=1609.22 GB/s, peak_mem=10.80 GB, model_size= 7.52 GB quant: int8wo, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization int8wo --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125123322, tok/s=131.13, mem/s= 553.61 GB/s, peak_mem= 6.81 GB, model_size= 4.22 GB quant: int4wo-64, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization int4wo-64 --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125123409, tok/s=168.82, mem/s=1267.56 GB/s, peak_mem=11.83 GB, model_size= 7.51 GB quant: float8wo, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization float8wo --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125123544, tok/s=141.95, mem/s=1065.36 GB/s, peak_mem=12.98 GB, model_size= 7.51 GB quant: float8dq-tensor, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization float8dq-tensor --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125123640, tok/s=142.23, mem/s=1067.49 GB/s, peak_mem=12.98 GB, model_size= 7.51 GB quant: float8dq-wo, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization float8dq-wo --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8

20241125130306, tok/s=163.37, mem/s=2452.20 GB/s, peak_mem=16.60 GB, model_size=15.01 GB quant: None, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125130348, tok/s=214.74, mem/s=1614.87 GB/s, peak_mem=10.80 GB, model_size= 7.52 GB quant: int8wo, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization int8wo --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125130423, tok/s=131.46, mem/s= 555.03 GB/s, peak_mem= 6.57 GB, model_size= 4.22 GB quant: int4wo-64, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization int4wo-64 --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125130601, tok/s=173.17, mem/s=1300.17 GB/s, peak_mem=11.83 GB, model_size= 7.51 GB quant: float8wo, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization float8wo --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125130727, tok/s=138.31, mem/s=1038.01 GB/s, peak_mem=12.98 GB, model_size= 7.51 GB quant: float8dq-tensor, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization float8dq-tensor --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
20241125130822, tok/s=140.46, mem/s=1054.20 GB/s, peak_mem=12.98 GB, model_size= 7.51 GB quant: float8dq-wo, mod: Meta-Llama-3.1-8B, kv_quant: False, compile: True, compile_prefill: False, dtype: torch.bfloat16, device: cuda repro: python generate.py --quantization float8dq-wo --checkpoint_path ../../../checkpoints/meta-llama/Meta-Llama-3.1-8B/model.pth --device cuda --precision torch.bfloat16 --compile --num_samples 5 --max_new_tokens 200 --batch_size 1 --top_k 200 --temperature 0.8
Loading