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

Fp8 testing #446

Closed
wants to merge 2 commits into from
Closed

Fp8 testing #446

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
aba1a10
add fp8 support
drisspg Jun 26, 2024
826bdaf
updates
drisspg Jul 4, 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
11 changes: 9 additions & 2 deletions scripts/hf_eval.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,7 +45,7 @@ def run_evaluation(repo_id, tasks, limit, device, precision, quantization, compi
model = AutoModelForCausalLM.from_pretrained(repo_id).to(device="cpu", dtype=precision)

if compile:
model = torch.compile(model, mode="max-autotune", fullgraph=True)
model = torch.compile(model, fullgraph=True)

if quantization == "int8dq":
change_linear_weights_to_int8_dqtensors(model)
Expand All @@ -56,6 +56,13 @@ def run_evaluation(repo_id, tasks, limit, device, precision, quantization, compi
change_linear_weights_to_int4_woqtensors(model.to(device=device))
elif quantization == "autoquant":
model = autoquant(model.to(device=device))
elif quantization == "fp8":
from float8_experimental.inference import quantize_to_float8, ActivationCasting, QuantConfig, ScalingGranularity
model.to(device)
quantize_to_float8(model, QuantConfig(ActivationCasting.DYNAMIC), scaling_granularity=ScalingGranularity.TensorWise)

pass # no quantization applied, model is already on device and precision dtype.

with torch.no_grad():
result = evaluate(
HFLM(
Expand All @@ -78,7 +85,7 @@ def run_evaluation(repo_id, tasks, limit, device, precision, quantization, compi
parser.add_argument('--limit', type=int, default=None, help='Number of eval samples to evaluate')
parser.add_argument('--precision', type=lambda x: getattr(torch, x.split(".")[-1]), default=torch.bfloat16, help='dtype precision to use')
parser.add_argument('--device', type=str, default="cuda", help='Device to use for evaluation')
parser.add_argument('-q', '--quantization', default = "None", choices=["int8dq", "int8wo", "int4wo","autoquant", "None"], help='Which quantization technique to apply')
parser.add_argument('-q', '--quantization', default = "None", choices=["int8dq", "int8wo", "int4wo","autoquant", "fp8", "None"], help='Which quantization technique to apply')
parser.add_argument('--compile', action='store_true', help='Whether to compile the model.')
parser.add_argument('--batch_size', type=int, default=1, help='Batch size to use for evaluation, note int8wo and int4wo work best with small batchsizes, int8dq works better with large batchsizes')
parser.add_argument('--max_length', type=int, default=None, help='Length of text to process at one time')
Expand Down
Loading