(act_bits=8, kv_bits=8, kv_group=64)
| 59 | |
| 60 | |
| 61 | def configure_deploy(act_bits=8, kv_bits=8, kv_group=64): |
| 62 | global ACT_BITS, KV_BITS, _KV_GROUP |
| 63 | kv_bits = 0 if int(kv_bits) >= 8 else int(kv_bits) |
| 64 | changed = (ACT_BITS, KV_BITS, _KV_GROUP) != (int(act_bits), int(kv_bits), int(kv_group)) |
| 65 | ACT_BITS, KV_BITS, _KV_GROUP = int(act_bits), int(kv_bits), int(kv_group) |
| 66 | if changed: |
| 67 | jax.clear_caches() |
| 68 | |
| 69 | |
| 70 | def quantize_params_configured(params): |
nothing calls this directly
no outgoing calls
no test coverage detected