833cc07d0e
dqn-production.toml batch_size=8192 was overriding GPU profile batch_size=64 on RTX 3050 via apply_to(), causing OOM/hang. GPU profile is the sole authority for batch_size — training profile controls learning params only. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
14 lines
270 B
TOML
14 lines
270 B
TOML
# RTX 3050 Ti (4GB VRAM) -- minimal footprint
|
|
[training]
|
|
batch_size = 64
|
|
num_atoms = 11
|
|
buffer_size = 0 # 0 = auto from VRAM
|
|
hidden_dim_base = 256
|
|
replay_buffer_vram_fraction = 0.40
|
|
|
|
[experience]
|
|
gpu_timesteps_per_episode = 100
|
|
|
|
[cuda]
|
|
cuda_stack_bytes = 16384 # 16KB
|