mirror of
https://github.com/harvard-edge/cs249r_book.git
synced 2026-07-21 00:23:30 -05:00
Vijay Janapa Reddi
fccee3c735
- Migrated all legacy constants from constants.py to hardware and model registries - Updated dozens of LEGO blocks in Volume 1 and Volume 2 to use Engine.solve and formulas - Resolved all resulting NameError, TypeError, and AttributeError exceptions in the inline Python blocks - Master Quarto HTML build now passes with zero execution failures
74 lines
2.8 KiB
Python
74 lines
2.8 KiB
Python
import re
|
|
import glob
|
|
|
|
def read(f):
|
|
with open(f, 'r') as file: return file.read()
|
|
def write(f, c):
|
|
with open(f, 'w') as file: file.write(c)
|
|
|
|
# Fix __init__.py to export formulas
|
|
init_path = 'mlsysim/mlsysim/__init__.py'
|
|
init_content = read(init_path)
|
|
if 'calc_transfer_time' not in init_content:
|
|
init_content = init_content.replace(
|
|
'from .fmt import fmt, check, MarkdownStr',
|
|
'from .core.formulas import *\nfrom .fmt import fmt, check, MarkdownStr'
|
|
)
|
|
write(init_path, init_content)
|
|
|
|
qmds = glob.glob('book/quarto/contents/**/*.qmd', recursive=True)
|
|
for q in qmds:
|
|
c = read(q)
|
|
orig = c
|
|
|
|
# 2. Fix Models.* aliases
|
|
models = ['GPT2', 'GPT3', 'GPT4', 'BERT_Base', 'BERT_Large', 'Llama2_70B', 'Llama2_7B', 'Llama3_8B', 'Llama3_70B']
|
|
for m in models:
|
|
c = re.sub(rf'\bModels\.{m}\b', f'Models.Language.{m}', c)
|
|
c = c.replace('Models.Language.Language.', 'Models.Language.')
|
|
c = c.replace('Models.Language.Vision.', 'Models.Vision.')
|
|
|
|
# 3. Fix ServingResult.weight_memory -> memory_footprint
|
|
c = c.replace('res.weight_memory.m_as', 'res.memory_footprint.m_as')
|
|
|
|
# 4. Fix iPhone15Pro15Pro
|
|
c = c.replace('iPhone15Pro15Pro', 'iPhone15Pro')
|
|
|
|
# 5. Fix constants.Hardware
|
|
c = c.replace('constants.Hardware', 'Hardware')
|
|
c = c.replace('constants.Models', 'Models')
|
|
|
|
# 6. Fix Hardware.Networks.Fabrics
|
|
c = c.replace('Hardware.Networks.Fabrics.', 'Hardware.Networks.')
|
|
|
|
# 8. Fix GenericNode
|
|
c = c.replace('Hardware.Cloud.GenericNode', 'Hardware.Cloud.GenericServer')
|
|
|
|
# 10. Fix Formatting Precision Error
|
|
c = c.replace('fmt(e_cpu_j / 50.0, precision=3, commas=False)', 'fmt(e_cpu_j / 50.0, precision=4, commas=False)')
|
|
|
|
# 13. Fix attention_heads -> heads
|
|
c = c.replace('.attention_heads', '.heads')
|
|
|
|
# 15. Fix IMAGENET_CLASSES
|
|
if 'IMAGENET_CLASSES' in c:
|
|
c = c.replace('IMAGENET_CLASSES', '1000') # Hardcode since it's 1000
|
|
|
|
# 16. Fix h100
|
|
if 'h100.tdp' in c:
|
|
c = c.replace('h100.tdp', 'h_h100.tdp')
|
|
c = c.replace('h100 = Hardware.Cloud.H100', 'h_h100 = Hardware.Cloud.H100')
|
|
|
|
# Fix calc_ring_allreduce_time
|
|
if 'calc_ring_allreduce_time(grad_70b, num_gpus, ib_ndr)' in c:
|
|
c = c.replace('calc_ring_allreduce_time(grad_70b, num_gpus, ib_ndr)', 'calc_ring_allreduce_time(grad_70b, num_gpus, ib_ndr, 0*ureg.ms)')
|
|
c = c.replace('calc_ring_allreduce_time(grad_1b, num_gpus, ib_ndr)', 'calc_ring_allreduce_time(grad_1b, num_gpus, ib_ndr, 0*ureg.ms)')
|
|
|
|
if 'calc_ring_allreduce_time(gradient_tensor, gpus_per_node, nvlink_bw)' in c:
|
|
c = c.replace('calc_ring_allreduce_time(gradient_tensor, gpus_per_node, nvlink_bw)', 'calc_ring_allreduce_time(gradient_tensor, gpus_per_node, nvlink_bw, 0*ureg.ms)')
|
|
|
|
if c != orig:
|
|
write(q, c)
|
|
print(f"Updated {q}")
|
|
|