basic learning
This commit is contained in:
parent
75ad46bfe9
commit
e2d29352c4
2 changed files with 391 additions and 184 deletions
359
main.lua
359
main.lua
|
|
@ -20,6 +20,63 @@ local function globalize(t)
|
|||
end
|
||||
end
|
||||
|
||||
-- configuration and globals.
|
||||
|
||||
math.randomseed(10)
|
||||
|
||||
local learning_rate = 1e-2
|
||||
local deviation = 2e-2
|
||||
|
||||
local epoch_i = 0
|
||||
local epoch_trials = 12
|
||||
local base_params
|
||||
local trial_i = 0
|
||||
local trial_noise = {}
|
||||
local trial_rewards = {}
|
||||
local trials_remaining = 0
|
||||
|
||||
local force_start = false
|
||||
local force_start_old = false
|
||||
|
||||
local enable_overlay = false
|
||||
local enable_network = true
|
||||
|
||||
local startsave = savestate.create(1)
|
||||
|
||||
local poketime = false
|
||||
|
||||
local sprite_input = {}
|
||||
local tile_input = {}
|
||||
|
||||
local reward
|
||||
local powerup_old
|
||||
local status_old
|
||||
local coins_old
|
||||
|
||||
local once = false
|
||||
local reset = true
|
||||
|
||||
local ok_routines = {
|
||||
[0x4] = true, -- sliding down flagpole
|
||||
[0x5] = true, -- end of level auto-walk
|
||||
[0x7] = true, -- start of level auto-walk
|
||||
[0x8] = true, -- normal (in control)
|
||||
[0x9] = true, -- acquiring mushroom
|
||||
[0xA] = true, -- losing big mario
|
||||
[0xB] = true, -- uhh
|
||||
[0xC] = true, -- acquiring fireflower
|
||||
}
|
||||
|
||||
local bad_states = {
|
||||
power = true,
|
||||
waiting_demo = true,
|
||||
playing_demo = true,
|
||||
unknown = true,
|
||||
lose = true,
|
||||
}
|
||||
|
||||
local state_old = ''
|
||||
|
||||
-- localize some stuff.
|
||||
|
||||
local print = print
|
||||
|
|
@ -51,6 +108,14 @@ local arshift = bit.arshift
|
|||
local rol = bit.rol
|
||||
local ror = bit.ror
|
||||
|
||||
-- utilities.
|
||||
|
||||
local function boolean_xor(a, b)
|
||||
if a and b then return false end
|
||||
if not a and not b then return false end
|
||||
return true
|
||||
end
|
||||
|
||||
local function clamp(x, l, u) return min(max(x, l), u) end
|
||||
|
||||
local function argmax(...)
|
||||
|
|
@ -70,6 +135,35 @@ local function argmax2(t)
|
|||
return t[1] > t[2]
|
||||
end
|
||||
|
||||
local function empty(t)
|
||||
for k, _ in pairs(t) do t[k] = nil end
|
||||
return t
|
||||
end
|
||||
|
||||
local function calc_mean_dev(x)
|
||||
local mean = 0
|
||||
for i, v in ipairs(x) do
|
||||
mean = mean + v / #x
|
||||
end
|
||||
|
||||
local dev = 0
|
||||
for i, v in ipairs(x) do
|
||||
local delta = v - mean
|
||||
dev = dev + delta * delta / #x
|
||||
end
|
||||
|
||||
return mean, dev
|
||||
end
|
||||
|
||||
local function normalize(x, out)
|
||||
out = out or x
|
||||
local mean, dev = calc_mean_dev(x)
|
||||
if dev <= 0 then dev = 1 end
|
||||
local devs = sqrt(dev)
|
||||
for i, v in ipairs(x) do out[i] = (v - mean) / devs end
|
||||
return out
|
||||
end
|
||||
|
||||
-- game-agnostic stuff (i.e. the network itself)
|
||||
|
||||
package.loaded['nn'] = nil -- DEBUG
|
||||
|
|
@ -100,9 +194,6 @@ end
|
|||
|
||||
-- and here we go with the game stuff.
|
||||
|
||||
local enable_overlay = false
|
||||
local enable_network = true
|
||||
|
||||
--[[
|
||||
https://gist.githubusercontent.com/1wErt3r/4048722/raw/59e88c0028a58c6d7b9156749230ccac647bc7d4/SMBDIS.ASM
|
||||
--]]
|
||||
|
|
@ -142,14 +233,17 @@ local rotation_offsets = { -- FIXME: not all of these are pixel-perfect.
|
|||
-8, -38,
|
||||
}
|
||||
|
||||
local startsave = savestate.create(1)
|
||||
local function get_timer()
|
||||
return R(0x7F8) * 100 + R(0x7F9) * 10 + R(0x7FA)
|
||||
end
|
||||
|
||||
local poketime = false
|
||||
local function set_timer(time)
|
||||
W(0x7F8, floor(time / 100))
|
||||
W(0x7F9, floor((time / 10) % 10))
|
||||
W(0x7FA, floor(time % 10))
|
||||
end
|
||||
|
||||
local sprite_input = {}
|
||||
local tile_input = {}
|
||||
|
||||
local function shitsprite(x, y, t)
|
||||
local function mark_sprite(x, y, t)
|
||||
if x < 0 or x >= 256 or y < 0 or y > 224 then
|
||||
sprite_input[#sprite_input+1] = 0
|
||||
sprite_input[#sprite_input+1] = 0
|
||||
|
|
@ -168,7 +262,7 @@ local function shitsprite(x, y, t)
|
|||
end
|
||||
end
|
||||
|
||||
local function shittile(x, y, t)
|
||||
local function mark_tile(x, y, t)
|
||||
tile_input[#tile_input+1] = t
|
||||
if t == 0 then return end
|
||||
if enable_overlay then
|
||||
|
|
@ -200,23 +294,9 @@ local function getxy(i, x_addr, y_addr, pageloc_addr, hipos_addr)
|
|||
return sx, sy
|
||||
end
|
||||
|
||||
local reward
|
||||
local powerup_old
|
||||
local status_old
|
||||
local coins_old
|
||||
|
||||
local once = false
|
||||
local reset = true
|
||||
|
||||
emu.poweron()
|
||||
emu.unpause()
|
||||
emu.speedmode("normal")
|
||||
|
||||
local function opermode() return R(0x770) end
|
||||
local function paused() return band(R(0x776), 1) end
|
||||
local function subroutine() return R(0xE) end
|
||||
|
||||
local function getstate()
|
||||
local function get_state()
|
||||
if R(0xE) == 0xFF then return 'power' end
|
||||
if R(0x774) > 0 then return 'lagging' end
|
||||
if R(0x7A2) > 0 then return 'waiting_demo' end
|
||||
|
|
@ -238,35 +318,14 @@ local function getstate()
|
|||
return 'unknown'
|
||||
end
|
||||
|
||||
local ok_routines = {
|
||||
[0x4] = true, -- sliding down flagpole
|
||||
[0x5] = true, -- end of level auto-walk
|
||||
[0x7] = true, -- start of level auto-walk
|
||||
[0x8] = true, -- normal (in control)
|
||||
[0x9] = true, -- acquiring mushroom
|
||||
[0xA] = true, -- losing big mario
|
||||
[0xB] = true, -- uhh
|
||||
[0xC] = true, -- acquiring fireflower
|
||||
}
|
||||
|
||||
local fuckstates = {
|
||||
power = true,
|
||||
waiting_demo = true,
|
||||
playing_demo = true,
|
||||
unknown = true,
|
||||
lose = true,
|
||||
paused = true,
|
||||
}
|
||||
|
||||
local function advance()
|
||||
emu.frameadvance()
|
||||
while emu.lagged() do emu.frameadvance() end -- skip lag frames.
|
||||
while R(0x774) > 0 do emu.frameadvance() end -- also lag frames.
|
||||
end
|
||||
|
||||
local state_old = ''
|
||||
while false do
|
||||
local state = getstate()
|
||||
local state = get_state()
|
||||
if state ~= state_old then
|
||||
print(emu.framecount(), state)
|
||||
state_old = state
|
||||
|
|
@ -301,9 +360,9 @@ local function handle_enemies()
|
|||
x, y = x + x_off, y + y_off
|
||||
end
|
||||
if invisible then
|
||||
shitsprite(0, 0, 0)
|
||||
mark_sprite(0, 0, 0)
|
||||
else
|
||||
shitsprite(x, y, tid + 1)
|
||||
mark_sprite(x, y, tid + 1)
|
||||
end
|
||||
end
|
||||
end
|
||||
|
|
@ -316,9 +375,9 @@ local function handle_fireballs()
|
|||
local state = R(0x24 + i)
|
||||
local invisible = state == 0
|
||||
if invisible then
|
||||
shitsprite(0, 0, 0)
|
||||
mark_sprite(0, 0, 0)
|
||||
else
|
||||
shitsprite(x, y, 257)
|
||||
mark_sprite(x, y, 257)
|
||||
end
|
||||
end
|
||||
end
|
||||
|
|
@ -330,9 +389,9 @@ local function handle_blocks()
|
|||
local state = R(0x26 + i)
|
||||
local invisible = state == 0
|
||||
if invisible then
|
||||
shitsprite(0, 0, 0)
|
||||
mark_sprite(0, 0, 0)
|
||||
else
|
||||
shitsprite(x, y, 258)
|
||||
mark_sprite(x, y, 258)
|
||||
end
|
||||
end
|
||||
end
|
||||
|
|
@ -347,9 +406,9 @@ local function handle_hammers()
|
|||
if state ~= 0
|
||||
and state >= 0x30
|
||||
then
|
||||
shitsprite(x, y, state + 1)
|
||||
mark_sprite(x, y, state + 1)
|
||||
else
|
||||
shitsprite(0, 0, 0)
|
||||
mark_sprite(0, 0, 0)
|
||||
end
|
||||
end
|
||||
end
|
||||
|
|
@ -360,9 +419,9 @@ local function handle_misc()
|
|||
x, y = x + 8, y + 8
|
||||
local state = R(0x33 + i)
|
||||
if state ~= 0 then
|
||||
shitsprite(x, y, state + 1)
|
||||
mark_sprite(x, y, state + 1)
|
||||
else
|
||||
shitsprite(0, 0, 0)
|
||||
mark_sprite(0, 0, 0)
|
||||
end
|
||||
end
|
||||
end
|
||||
|
|
@ -383,25 +442,104 @@ local function handle_tiles()
|
|||
end
|
||||
local sx = x * 16 + 8 - tile_scroll_remainder
|
||||
local sy = y * 16 + 40
|
||||
shittile(sx, sy, t)
|
||||
mark_tile(sx, sy, t)
|
||||
end
|
||||
end
|
||||
end
|
||||
|
||||
local function doreset()
|
||||
print("resetting in state:", getstate())
|
||||
local function learn_from_epoch()
|
||||
print()
|
||||
print('rewards:', trial_rewards)
|
||||
normalize(trial_rewards)
|
||||
print('normalized:', trial_rewards)
|
||||
|
||||
local reward_mean, reward_dev = calc_mean_dev(trial_rewards)
|
||||
|
||||
local step = nn.zeros(#base_params)
|
||||
for i = 1, epoch_trials do
|
||||
local reward = trial_rewards[i]
|
||||
local noise = trial_noise[i]
|
||||
for j, v in ipairs(noise) do
|
||||
step[j] = step[j] + reward * v
|
||||
end
|
||||
end
|
||||
|
||||
local magnitude = learning_rate / deviation
|
||||
print('stepping with magnitude', magnitude)
|
||||
-- throw the division from the averaging in there too.
|
||||
local altogether = magnitude / epoch_trials
|
||||
for i, v in ipairs(step) do
|
||||
step[i] = altogether * v
|
||||
end
|
||||
|
||||
local step_mean, step_dev = calc_mean_dev(step)
|
||||
print("step mean:", step_mean)
|
||||
print("step stddev:", step_dev)
|
||||
|
||||
if step_dev > 1e-8 then
|
||||
for i, v in ipairs(base_params) do
|
||||
base_params[i] = v + step[i]
|
||||
end
|
||||
else
|
||||
-- we didn't get anywhere. step in a random direction.
|
||||
local noise = trial_noise[1]
|
||||
for i, v in ipairs(base_params) do
|
||||
base_params[i] = v + magnitude * noise[i]
|
||||
end
|
||||
end
|
||||
|
||||
network:distribute(base_params)
|
||||
|
||||
network:save()
|
||||
|
||||
print()
|
||||
end
|
||||
|
||||
local function prepare_epoch()
|
||||
print('preparing epoch '..tostring(epoch_i)..'. this might take a while.')
|
||||
base_params = network:collect()
|
||||
empty(trial_noise)
|
||||
empty(trial_rewards)
|
||||
for i = 1, epoch_trials do
|
||||
local noise = nn.zeros(#base_params)
|
||||
for j = 1, #base_params do noise[j] = nn.normal() end
|
||||
trial_noise[i] = noise
|
||||
end
|
||||
trial_i = 0
|
||||
end
|
||||
|
||||
local function load_next_trial()
|
||||
trial_i = trial_i + 1
|
||||
print('loading trial', trial_i)
|
||||
local W = nn.copy(base_params)
|
||||
local noise = trial_noise[trial_i]
|
||||
for i, v in ipairs(base_params) do
|
||||
W[i] = v + deviation * noise[i]
|
||||
end
|
||||
network:distribute(W)
|
||||
end
|
||||
|
||||
local function do_reset()
|
||||
print("resetting in state:", get_state())
|
||||
|
||||
if trial_i > 0 then trial_rewards[trial_i] = reward end
|
||||
|
||||
if epoch_i == 0 or trial_i == epoch_trials then
|
||||
if epoch_i > 0 then learn_from_epoch() else network:reset() end
|
||||
epoch_i = epoch_i + 1
|
||||
prepare_epoch()
|
||||
end
|
||||
|
||||
if once then
|
||||
savestate.load(startsave)
|
||||
print("end of trial reward:", reward)
|
||||
print()
|
||||
else
|
||||
savestate.save(startsave)
|
||||
end
|
||||
once = true
|
||||
|
||||
-- bit of a hack:
|
||||
if getstate() == 'loading' then advance() end
|
||||
if get_state() == 'loading' then advance() end
|
||||
reward = 0
|
||||
powerup_old = R(0x754)
|
||||
status_old = R(0x756)
|
||||
|
|
@ -413,13 +551,29 @@ local function doreset()
|
|||
|
||||
emu.frameadvance() -- prevents emulator from quirking up.
|
||||
|
||||
print()
|
||||
load_next_trial()
|
||||
|
||||
reset = false
|
||||
if network ~= nil then network:reset() end -- FIXME: hack
|
||||
end
|
||||
|
||||
local function init()
|
||||
network = make_network(279, 8)
|
||||
network:reset()
|
||||
print("parameters:", network.n_param)
|
||||
|
||||
emu.poweron()
|
||||
emu.unpause()
|
||||
emu.speedmode("normal")
|
||||
end
|
||||
|
||||
init()
|
||||
|
||||
while true do
|
||||
while fuckstates[getstate()] do
|
||||
--gui.text(120, 124, ("%02X"):format(subroutine()), '#FFFFFF', '#0000003F')
|
||||
gui.text(4, 12, get_state(), '#FFFFFF', '#0000003F')
|
||||
|
||||
while bad_states[get_state()] do
|
||||
--gui.text(120, 124, ("%02X"):format(R(0xE)), '#FFFFFF', '#0000003F')
|
||||
-- mash the start button until we have control.
|
||||
-- TODO: learn this too.
|
||||
--local jp = joypad.read(1)
|
||||
|
|
@ -437,44 +591,39 @@ while true do
|
|||
|
||||
reset = true
|
||||
|
||||
gui.text(4, 12, getstate(), '#FFFFFF', '#0000003F')
|
||||
advance()
|
||||
gui.text(4, 12, get_state(), '#FFFFFF', '#0000003F')
|
||||
|
||||
-- bit of a hack:
|
||||
while getstate() == "loading" do advance() end
|
||||
state_old = getstate()
|
||||
while get_state() == "loading" do advance() end
|
||||
state_old = get_state()
|
||||
end
|
||||
|
||||
if reset then doreset() end
|
||||
if reset then do_reset() end
|
||||
|
||||
if not enable_network then
|
||||
-- infinite time cheat. super handy for testing.
|
||||
if R(0xE) == 8 then
|
||||
W(0x7F8, 9)
|
||||
W(0x7F9, 9)
|
||||
W(0x7FA, 10)
|
||||
set_timer(667)
|
||||
poketime = true
|
||||
elseif poketime then
|
||||
poketime = false
|
||||
W(0x7F8, 0)
|
||||
W(0x7F9, 0)
|
||||
W(0x7FA, 1)
|
||||
set_timer(1)
|
||||
end
|
||||
|
||||
-- infinite lives.
|
||||
W(0x75A, 1)
|
||||
end
|
||||
|
||||
-- empty input lists without creating a new table.
|
||||
for k, v in pairs(sprite_input) do sprite_input[k] = nil end
|
||||
for k, v in pairs(tile_input) do tile_input[k] = nil end
|
||||
empty(sprite_input)
|
||||
empty(tile_input)
|
||||
|
||||
-- player
|
||||
-- TODO: add check if mario is playable.
|
||||
-- TODO: check if mario is playable.
|
||||
local x, y = getxy(0, 0x86, 0xCE, 0x6D, 0xB5)
|
||||
local powerup = R(0x754)
|
||||
local status = R(0x756)
|
||||
shitsprite(x + 8, y + 24, -powerup - 1)
|
||||
mark_sprite(x + 8, y + 24, -powerup - 1)
|
||||
|
||||
handle_enemies()
|
||||
handle_fireballs()
|
||||
|
|
@ -484,6 +633,8 @@ while true do
|
|||
handle_misc()
|
||||
handle_tiles()
|
||||
|
||||
local ingame_paused = get_state() == "paused"
|
||||
|
||||
local coins = R(0x7ED) * 10 + R(0x7EE)
|
||||
local coins_delta = coins - coins_old
|
||||
-- handle wrap-around.
|
||||
|
|
@ -493,45 +644,48 @@ while true do
|
|||
-- 2 is fire mario.
|
||||
local status_delta = clamp(status - status_old, -1, 1)
|
||||
local screen_scroll_delta = R(0x775)
|
||||
local flagpole_bonus = subroutine() == 4 and 1 or 0
|
||||
local flagpole_bonus = R(0xE) == 4 and 1 or 0
|
||||
local reward_delta = screen_scroll_delta + status_delta * 256 + flagpole_bonus
|
||||
|
||||
reward = reward + reward_delta
|
||||
if not ingame_paused then reward = reward + reward_delta end
|
||||
|
||||
--gui.text(4, 12, ("%02X"):format(#sprite_input), '#FFFFFF', '#0000003F')
|
||||
--gui.text(4, 22, ("%02X"):format(#tile_input), '#FFFFFF', '#0000003F')
|
||||
gui.text(72, 12, ("%+4i"):format(reward_delta), '#FFFFFF', '#0000003F')
|
||||
gui.text(112, 12, ("%+4i"):format(reward), '#FFFFFF', '#0000003F')
|
||||
|
||||
if getstate() == 'dead' and state_old ~= 'dead' then
|
||||
if get_state() == 'dead' and state_old ~= 'dead' then
|
||||
print("dead. lives remaining:", R(0x75A, 0))
|
||||
if R(0x75A, 0) == 0 then reset = true end
|
||||
end
|
||||
if getstate() == 'lose' then
|
||||
if get_state() == 'lose' then
|
||||
print("ran out of lives.")
|
||||
reset = true
|
||||
end
|
||||
|
||||
-- lose a point for every frame paused.
|
||||
if ingame_paused then reward = reward - 1 end
|
||||
|
||||
-- every few frames mario stands still, forcibly decrease the timer.
|
||||
-- this includes having the game paused.
|
||||
-- TODO: more robust. doesn't detect moonwalking against a wall.
|
||||
local timer = get_timer()
|
||||
local timer_loser = 1/5
|
||||
if math.random() > 1 - timer_loser and R(0x1D) == 0 and R(0x57) == 0 then
|
||||
local timer = R(0x7F8) * 100 + R(0x7F9) * 10 + R(0x7FA)
|
||||
if ingame_paused or math.random() > 1 - timer_loser and R(0x1D) == 0 and R(0x57) == 0 then
|
||||
timer = clamp(timer - 1, 0, 400)
|
||||
W(0x7F8, floor(timer / 100))
|
||||
W(0x7F9, floor((timer / 10) % 10))
|
||||
W(0x7FA, floor(timer % 10))
|
||||
set_timer(timer)
|
||||
end
|
||||
|
||||
-- if we've run out of time while the game is paused...
|
||||
-- that's cheating! unpause.
|
||||
force_start = ingame_paused and timer == 0
|
||||
|
||||
local X = {} -- TODO: cache.
|
||||
for i, v in ipairs(sprite_input) do insert(X, v / 256) end
|
||||
for i, v in ipairs(tile_input) do insert(X, v / 256) end
|
||||
--error(#X)
|
||||
|
||||
if network == nil then
|
||||
network = make_network(#X, 8); network:reset()
|
||||
print("parameters:", network.n_param)
|
||||
end
|
||||
|
||||
if enable_network and getstate() == 'playing' then
|
||||
if enable_network and get_state() == 'playing' or ingame_paused then
|
||||
local outputs = network:forward(X)
|
||||
local softmaxed = {
|
||||
outputs[nn_z[1]],
|
||||
|
|
@ -553,12 +707,25 @@ while true do
|
|||
start = argmax2(softmaxed[7]),
|
||||
select = argmax2(softmaxed[8]),
|
||||
}
|
||||
if force_start then
|
||||
jp = {
|
||||
up = false,
|
||||
down = false,
|
||||
left = false,
|
||||
right = false,
|
||||
A = false,
|
||||
B = false,
|
||||
start = force_start_old,
|
||||
select = false,
|
||||
}
|
||||
end
|
||||
joypad.write(1, jp)
|
||||
end
|
||||
|
||||
coins_old = coins
|
||||
powerup_old = powerup
|
||||
status_old = status
|
||||
state_old = getstate()
|
||||
force_start_old = force_start
|
||||
state_old = get_state()
|
||||
advance()
|
||||
end
|
||||
|
|
|
|||
206
nn.lua
206
nn.lua
|
|
@ -13,6 +13,7 @@ local cos = math.cos
|
|||
local sin = math.sin
|
||||
local insert = table.insert
|
||||
local remove = table.remove
|
||||
local open = io.open
|
||||
|
||||
local bor = bit.bor
|
||||
|
||||
|
|
@ -24,6 +25,17 @@ local function contains(t, a)
|
|||
return false
|
||||
end
|
||||
|
||||
local function prod(x, ...)
|
||||
if type(x) == "table" then
|
||||
return prod(unpack(x))
|
||||
end
|
||||
local ret = x
|
||||
for i = 1, select("#", ...) do
|
||||
ret = ret * select(i, ...)
|
||||
end
|
||||
return ret
|
||||
end
|
||||
|
||||
local function normal() -- box muller
|
||||
return sqrt(-2 * log(uniform() + 1e-8) + 1e-8) * cos(2 * pi * uniform())
|
||||
end
|
||||
|
|
@ -58,28 +70,53 @@ local function copy(t) -- shallow copy
|
|||
end
|
||||
|
||||
local function allocate(t, out, init)
|
||||
-- FIXME: this code is fucking disgusting.
|
||||
|
||||
out = out or {}
|
||||
assert(type(out) == "table", type(out))
|
||||
if type(t) == "number" then
|
||||
local size = t
|
||||
if init ~= nil then
|
||||
return init(zeros(size, out))
|
||||
else
|
||||
return zeros(size, out)
|
||||
end
|
||||
end
|
||||
|
||||
local function levelorder(field, node_in, nodes)
|
||||
-- horribly inefficient.
|
||||
nodes = nodes or {}
|
||||
local q = {node_in}
|
||||
while #q > 0 do
|
||||
local node = q[1]
|
||||
remove(q, 1)
|
||||
insert(nodes, node)
|
||||
for _, child in ipairs(node[field]) do
|
||||
q[#q+1] = child
|
||||
end
|
||||
local topsize = t[1]
|
||||
t = copy(t)
|
||||
remove(t, 1)
|
||||
if #t == 1 then t = t[1] end
|
||||
for i = 1, topsize do
|
||||
local res = allocate(t, nil, init)
|
||||
assert(res ~= nil)
|
||||
insert(out, res)
|
||||
end
|
||||
return out
|
||||
return nodes
|
||||
end
|
||||
|
||||
local function traverse(node_in, node_out, nodes)
|
||||
nodes = nodes or {}
|
||||
local down = levelorder('children', node_in, {})
|
||||
local up = levelorder('parents', node_out, {})
|
||||
local seen = {}
|
||||
for _, node in ipairs(up) do
|
||||
seen[node] = bor(seen[node] or 0, 1)
|
||||
end
|
||||
for _, node in ipairs(down) do
|
||||
seen[node] = bor(seen[node] or 0, 2)
|
||||
if seen[node] == 3 then
|
||||
insert(nodes, node)
|
||||
end
|
||||
end
|
||||
return nodes
|
||||
end
|
||||
|
||||
local function traverse_all(nodes_in, nodes_out, nodes)
|
||||
local all_in = {children={}}
|
||||
local all_out = {parents={}}
|
||||
for _, node in ipairs(nodes_in) do insert(all_in.children, node) end
|
||||
for _, node in ipairs(nodes_out) do insert(all_out.parents, node) end
|
||||
return traverse(all_in, all_out, nodes or {})
|
||||
end
|
||||
|
||||
local Weights = Base:extend()
|
||||
|
|
@ -95,24 +132,13 @@ function Weights:init(weight_init)
|
|||
end
|
||||
|
||||
function Weights:allocate(fan_in, fan_out)
|
||||
self.size = prod(self.shape)
|
||||
return allocate(self.size, self, function(t)
|
||||
--print('initializing weights of size', self.size, 'with fans', fan_in, fan_out)
|
||||
return self.weight_init(t, fan_in, fan_out)
|
||||
end)
|
||||
end
|
||||
|
||||
--[[
|
||||
local w = Weights(init_he_uniform)
|
||||
w.size = {16, 16}
|
||||
w:allocate(16, 16)
|
||||
print(w)
|
||||
do return end
|
||||
|
||||
local w = zeros(16)
|
||||
for i = 1, #w do w[i] = normal() * 1920 / 2560 end
|
||||
print(w)
|
||||
--]]
|
||||
|
||||
local counter = {}
|
||||
function Layer:init(name)
|
||||
assert(type(name) == "string")
|
||||
|
|
@ -152,18 +178,7 @@ function Layer:_new_weights(init)
|
|||
return w
|
||||
end
|
||||
|
||||
local function prod(x, ...)
|
||||
if type(x) == "table" then
|
||||
return prod(unpack(x))
|
||||
end
|
||||
local ret = x
|
||||
for i = 1, select("#", ...) do
|
||||
ret = ret * select(i, ...)
|
||||
end
|
||||
return ret
|
||||
end
|
||||
|
||||
function Layer:getsize()
|
||||
function Layer:get_size()
|
||||
local size = 0
|
||||
for i, w in ipairs(self.weights) do size = size + prod(w.size) end
|
||||
return size
|
||||
|
|
@ -237,8 +252,8 @@ end
|
|||
|
||||
function Dense:make_shape(parent)
|
||||
self.size_in = parent.size_out
|
||||
self.coeffs.size = {self.dim, self.size_in}
|
||||
self.biases.size = self.dim
|
||||
self.coeffs.shape = {self.size_in, self.dim}
|
||||
self.biases.shape = self.dim
|
||||
end
|
||||
|
||||
function Dense:forward(X)
|
||||
|
|
@ -246,11 +261,11 @@ function Dense:forward(X)
|
|||
self.cache = self.cache or zeros(self.size_out)
|
||||
local Y = self.cache
|
||||
|
||||
for i = 1, #self.coeffs do
|
||||
for i = 1, self.dim do
|
||||
local res = 0
|
||||
local c = self.coeffs[i]
|
||||
local c = (i - 1) * #X
|
||||
for j = 1, #X do
|
||||
res = res + X[j] * c[j]
|
||||
res = res + X[j] * self.coeffs[c + j]
|
||||
end
|
||||
Y[i] = res + self.biases[i]
|
||||
end
|
||||
|
|
@ -281,46 +296,6 @@ function Softmax:forward(X)
|
|||
return Y
|
||||
end
|
||||
|
||||
local function levelorder(field, node_in, nodes)
|
||||
-- horribly inefficient.
|
||||
nodes = nodes or {}
|
||||
local q = {node_in}
|
||||
while #q > 0 do
|
||||
local node = q[1]
|
||||
remove(q, 1)
|
||||
insert(nodes, node)
|
||||
for _, child in ipairs(node[field]) do
|
||||
q[#q+1] = child
|
||||
end
|
||||
end
|
||||
return nodes
|
||||
end
|
||||
|
||||
local function traverse(node_in, node_out, nodes)
|
||||
nodes = nodes or {}
|
||||
local down = levelorder('children', node_in, {})
|
||||
local up = levelorder('parents', node_out, {})
|
||||
local seen = {}
|
||||
for _, node in ipairs(up) do
|
||||
seen[node] = bor(seen[node] or 0, 1)
|
||||
end
|
||||
for _, node in ipairs(down) do
|
||||
seen[node] = bor(seen[node] or 0, 2)
|
||||
if seen[node] == 3 then
|
||||
insert(nodes, node)
|
||||
end
|
||||
end
|
||||
return nodes
|
||||
end
|
||||
|
||||
local function traverse_all(nodes_in, nodes_out, nodes)
|
||||
local all_in = {children={}}
|
||||
local all_out = {parents={}}
|
||||
for _, node in ipairs(nodes_in) do insert(all_in.children, node) end
|
||||
for _, node in ipairs(nodes_out) do insert(all_out.parents, node) end
|
||||
return traverse(all_in, all_out, nodes or {})
|
||||
end
|
||||
|
||||
function Model:init(nodes_in, nodes_out)
|
||||
assert(#nodes_in > 0, #nodes_in)
|
||||
assert(#nodes_out > 0, #nodes_out)
|
||||
|
|
@ -338,7 +313,7 @@ function Model:reset()
|
|||
self.n_param = 0
|
||||
for _, node in ipairs(self.nodes) do
|
||||
node:init_weights()
|
||||
self.n_param = self.n_param + node:getsize()
|
||||
self.n_param = self.n_param + node:get_size()
|
||||
end
|
||||
end
|
||||
|
||||
|
|
@ -359,10 +334,75 @@ function Model:forward(X)
|
|||
return outputs
|
||||
end
|
||||
|
||||
function Model:collect()
|
||||
-- return a flat array of all the weights in the graph.
|
||||
-- if Lua had slices, we wouldn't need this. future library idea?
|
||||
assert(self.n_param >= 0, self.n_param)
|
||||
local W = zeros(self.n_param)
|
||||
local i = 0
|
||||
for _, node in ipairs(self.nodes) do
|
||||
for _, w in ipairs(node.weights) do
|
||||
for j, v in ipairs(w) do
|
||||
W[i+j] = v
|
||||
end
|
||||
i = i + #w
|
||||
end
|
||||
end
|
||||
return W
|
||||
end
|
||||
|
||||
function Model:distribute(W)
|
||||
-- inverse operation of collect().
|
||||
local i = 0
|
||||
for _, node in ipairs(self.nodes) do
|
||||
for _, w in ipairs(node.weights) do
|
||||
for j, v in ipairs(w) do
|
||||
w[j] = W[i+j]
|
||||
end
|
||||
i = i + #w
|
||||
end
|
||||
end
|
||||
end
|
||||
|
||||
function Model:save(fn)
|
||||
local fn = fn or 'network.txt'
|
||||
local f = open(fn, 'w')
|
||||
if f == nil then error("Failed to save network to file "..fn) end
|
||||
local W = self:collect()
|
||||
for i, v in ipairs(W) do
|
||||
f:write(v)
|
||||
f:write('\n')
|
||||
end
|
||||
f:close()
|
||||
end
|
||||
|
||||
function Model:load(fn)
|
||||
local fn = fn or 'network.txt'
|
||||
local f = open(fn, 'r')
|
||||
if f == nil then
|
||||
error("Failed to load network from file "..fn)
|
||||
end
|
||||
|
||||
local W = zeros(self.n_param)
|
||||
local i = 0
|
||||
for line in f:lines() do
|
||||
i = i + 1
|
||||
local n = tonumber(line)
|
||||
if n == nil then
|
||||
error("Failed reading line "..tostring(i).." of file "..fn)
|
||||
end
|
||||
W[i] = n
|
||||
end
|
||||
f:close()
|
||||
|
||||
self:distribute(W)
|
||||
end
|
||||
|
||||
return {
|
||||
uniform = uniform,
|
||||
normal = normal,
|
||||
|
||||
copy = copy,
|
||||
zeros = zeros,
|
||||
init_zeros = init_zeros,
|
||||
init_he_uniform = init_he_uniform,
|
||||
|
|
|
|||
Loading…
Reference in a new issue