diff options
| -rw-r--r-- | Makefile | 48 | ||||
| -rw-r--r-- | bitsandbytes/functional.py | 4 | ||||
| -rw-r--r-- | bitsandbytes/optim/optimizer.py | 8 | ||||
| -rw-r--r-- | csrc/kernels.cu | 122 | ||||
| -rw-r--r-- | csrc/pythonInterface.c | 2 |
5 files changed, 102 insertions, 82 deletions
@@ -15,29 +15,31 @@ INCLUDE := -I $(CUDA_HOME)/include -I $(ROOT_DIR)/csrc -I $(CONDA_PREFIX)/inclu LIB := -L $(CUDA_HOME)/lib64 -lcudart -lcuda -lcublas -lcurand -lcusparse -L $(CONDA_PREFIX)/lib # NVIDIA NVCC compilation flags -COMPUTE_CAPABILITY := -gencode arch=compute_35,code=sm_35 # Kepler -COMPUTE_CAPABILITY += -gencode arch=compute_37,code=sm_37 # Kepler -COMPUTE_CAPABILITY += -gencode arch=compute_50,code=sm_50 # Maxwell -COMPUTE_CAPABILITY += -gencode arch=compute_52,code=sm_52 # Maxwell -COMPUTE_CAPABILITY += -gencode arch=compute_60,code=sm_60 # Pascal -COMPUTE_CAPABILITY += -gencode arch=compute_61,code=sm_61 # Pascal -COMPUTE_CAPABILITY += -gencode arch=compute_70,code=sm_70 # Volta -COMPUTE_CAPABILITY += -gencode arch=compute_72,code=sm_72 # Volta -COMPUTE_CAPABILITY += -gencode arch=compute_72,code=sm_72 # Volta - -# CUDA 9.2 supports CC 3.0, but CUDA >= 11.0 does not -CC_CUDA92 := -gencode arch=compute_30,code=sm_30 - -# Later versions of CUDA support the new architectures -CC_CUDA10x := -gencode arch=compute_30,code=sm_30 -CC_CUDA10x += -gencode arch=compute_75,code=sm_75 - -CC_CUDA110 := -gencode arch=compute_75,code=sm_75 -CC_CUDA110 += -gencode arch=compute_80,code=sm_80 - -CC_CUDA11x := -gencode arch=compute_75,code=sm_75 -CC_CUDA11x += -gencode arch=compute_80,code=sm_80 -CC_CUDA11x += -gencode arch=compute_86,code=sm_86 +#COMPUTE_CAPABILITY := -gencode arch=compute_35,code=sm_35 # Kepler +#COMPUTE_CAPABILITY += -gencode arch=compute_37,code=sm_37 # Kepler +#COMPUTE_CAPABILITY += -gencode arch=compute_50,code=sm_50 # Maxwell +#COMPUTE_CAPABILITY += -gencode arch=compute_52,code=sm_52 # Maxwell +#COMPUTE_CAPABILITY += -gencode arch=compute_60,code=sm_60 # Pascal +#COMPUTE_CAPABILITY += -gencode arch=compute_61,code=sm_61 # Pascal +#COMPUTE_CAPABILITY += -gencode arch=compute_70,code=sm_70 # Volta +#COMPUTE_CAPABILITY += -gencode arch=compute_72,code=sm_72 # Volta +#COMPUTE_CAPABILITY += -gencode arch=compute_72,code=sm_72 # Volta +# +## CUDA 9.2 supports CC 3.0, but CUDA >= 11.0 does not +#CC_CUDA92 := -gencode arch=compute_30,code=sm_30 +# +## Later versions of CUDA support the new architectures +#CC_CUDA10x := -gencode arch=compute_30,code=sm_30 +#CC_CUDA10x += -gencode arch=compute_75,code=sm_75 +# +#CC_CUDA110 := -gencode arch=compute_75,code=sm_75 +#CC_CUDA110 += -gencode arch=compute_80,code=sm_80 +# +#CC_CUDA11x := -gencode arch=compute_75,code=sm_75 +#CC_CUDA11x += -gencode arch=compute_80,code=sm_80 +#CC_CUDA11x += -gencode arch=compute_86,code=sm_86 + +COMPUTE_CAPABILITY := -gencode arch=compute_70,code=sm_70 # Volta all: $(ROOT_DIR)/dependencies/cub $(BUILD_DIR) diff --git a/bitsandbytes/functional.py b/bitsandbytes/functional.py index 48ab40c..9fe1345 100644 --- a/bitsandbytes/functional.py +++ b/bitsandbytes/functional.py @@ -486,13 +486,13 @@ def optimizer_update_8bit_blockwise(optimizer_name: str, g: Tensor, p: Tensor, s str2optimizer8bit_blockwise[optimizer_name][0](get_ptr(p), get_ptr(g), get_ptr(state1), get_ptr(state2), ct.c_float(beta1), ct.c_float(beta2), ct.c_float(eps), ct.c_int32(step), ct.c_float(lr), get_ptr(qmap1), get_ptr(qmap2), - get_ptr(absmax1), get_ptr(absmax2), ct.c_float(weight_decay), ct.c_float(gnorm_scale), + get_ptr(absmax1), get_ptr(absmax2), ct.c_float(weight_decay), ct.c_float(gnorm_scale), ct.c_bool(skip_zeros), ct.c_int32(g.numel())) elif g.dtype == torch.float16 and state1.dtype == torch.uint8: str2optimizer8bit_blockwise[optimizer_name][1](get_ptr(p), get_ptr(g), get_ptr(state1), get_ptr(state2), ct.c_float(beta1), ct.c_float(beta2), ct.c_float(eps), ct.c_int32(step), ct.c_float(lr), get_ptr(qmap1), get_ptr(qmap2), - get_ptr(absmax1), get_ptr(absmax2), ct.c_float(weight_decay), ct.c_float(gnorm_scale), + get_ptr(absmax1), get_ptr(absmax2), ct.c_float(weight_decay), ct.c_float(gnorm_scale), ct.c_bool(skip_zeros), ct.c_int32(g.numel())) else: raise ValueError(f'Gradient+optimizer bit data type combination not supported: grad {g.dtype}, optimizer {state1.dtype}') diff --git a/bitsandbytes/optim/optimizer.py b/bitsandbytes/optim/optimizer.py index 25512b1..4b70b5c 100644 --- a/bitsandbytes/optim/optimizer.py +++ b/bitsandbytes/optim/optimizer.py @@ -336,7 +336,7 @@ class Optimizer2State(Optimizer8bit): if state['state1'].dtype == torch.float: F.optimizer_update_32bit(self.optimizer_name, grad, p, state['state1'], config['betas'][0], config['eps'], step, config['lr'], state['state2'], config['betas'][1], config['weight_decay'], gnorm_scale, - state['unorm_vec'] if config['max_unorm'] > 0.0 else None, max_unorm=config['max_unorm']) + state['unorm_vec'] if config['max_unorm'] > 0.0 else None, max_unorm=config['max_unorm'], skip_zeros=config['skip_zeros']) elif state['state1'].dtype == torch.uint8 and not config['block_wise']: F.optimizer_update_8bit(self.optimizer_name, grad, p, state['state1'], state['state2'], config['betas'][0], config['betas'][1], @@ -352,7 +352,7 @@ class Optimizer2State(Optimizer8bit): F.optimizer_update_8bit_blockwise(self.optimizer_name, grad, p, state['state1'], state['state2'], config['betas'][0], config['betas'][1], config['eps'], step, config['lr'], state['qmap1'], state['qmap2'], state['absmax1'], state['absmax2'], - config['weight_decay'], gnorm_scale=gnorm_scale) + config['weight_decay'], gnorm_scale=gnorm_scale, skip_zeros=config['skip_zeros']) class Optimizer1State(Optimizer8bit): @@ -450,7 +450,7 @@ class Optimizer1State(Optimizer8bit): F.optimizer_update_32bit(self.optimizer_name, grad, p, state['state1'], config['betas'][0], config['eps'], step, config['lr'], None, 0.0, config['weight_decay'], gnorm_scale, state['unorm_vec'] if config['max_unorm'] > 0.0 else None, max_unorm=config['max_unorm'], - skip_zeros=False) + skip_zeros=config['skip_zeros']) elif state['state1'].dtype == torch.uint8 and not config['block_wise']: F.optimizer_update_8bit(self.optimizer_name, grad, p, state['state1'], None, config['betas'][0], config['betas'][1], @@ -463,4 +463,4 @@ class Optimizer1State(Optimizer8bit): F.optimizer_update_8bit_blockwise(self.optimizer_name, grad, p, state['state1'], None, config['betas'][0], config['betas'][1], config['eps'], step, config['lr'], state['qmap1'], None, state['absmax1'], None, - config['weight_decay'], gnorm_scale=gnorm_scale, skip_zeros=False) + config['weight_decay'], gnorm_scale=gnorm_scale, skip_zeros=config['skip_zeros']) diff --git a/csrc/kernels.cu b/csrc/kernels.cu index f8f7b62..d1fa253 100644 --- a/csrc/kernels.cu +++ b/csrc/kernels.cu @@ -715,9 +715,12 @@ __global__ void kOptimizer32bit2State(T* g, T* p, switch(OPTIMIZER) { case ADAM: - s1_vals[j] = s1_vals[j]*beta1 + ((1.0f -beta1)*((float)g_vals[j])); - s2_vals[j] = s2_vals[j]*beta2 + ((1.0f -beta2)*(((float)g_vals[j])*((float)g_vals[j]))); - p_vals[j] = ((float)p_vals[j]) + (update_scale*step_size*(s1_vals[j]/(sqrtf(s2_vals[j])+(eps*correction2)))); + if(!skip_zeros || (skip_zeros && g_vals[j] != (T)0.0)) + { + s1_vals[j] = s1_vals[j]*beta1 + ((1.0f -beta1)*((float)g_vals[j])); + s2_vals[j] = s2_vals[j]*beta2 + ((1.0f -beta2)*(((float)g_vals[j])*((float)g_vals[j]))); + p_vals[j] = ((float)p_vals[j]) + (update_scale*step_size*(s1_vals[j]/(sqrtf(s2_vals[j])+(eps*correction2)))); + } break; } } @@ -865,21 +868,24 @@ __global__ void kOptimizer32bit1State(T *g, T *p, # pragma unroll 4 for(unsigned int j = 0; j < NUM_PER_THREAD; j++) { - switch(OPTIMIZER) - { - case MOMENTUM: - if(step == 1) - s1_vals[j] = (float)g_vals[j]; - else - s1_vals[j] = s1_vals[j]*beta1 + ((float)g_vals[j]); - - p_vals[j] = ((float)p_vals[j]) + update_scale*(-lr*(s1_vals[j])); - break; - case RMSPROP: - s1_vals[j] = s1_vals[j]*beta1 + ((1.0f-beta1)*((float)g_vals[j])*((float)g_vals[j])); - p_vals[j] = ((float)p_vals[j]) - update_scale*(lr*__fdividef((float)g_vals[j],sqrtf((float)s1_vals[j])+eps)); - break; - } + if(!skip_zeros || (skip_zeros && g_vals[j] != (T)0.0)) + { + switch(OPTIMIZER) + { + case MOMENTUM: + if(step == 1) + s1_vals[j] = (float)g_vals[j]; + else + s1_vals[j] = s1_vals[j]*beta1 + ((float)g_vals[j]); + + p_vals[j] = ((float)p_vals[j]) + update_scale*(-lr*(s1_vals[j])); + break; + case RMSPROP: + s1_vals[j] = s1_vals[j]*beta1 + ((1.0f-beta1)*((float)g_vals[j])*((float)g_vals[j])); + p_vals[j] = ((float)p_vals[j]) - update_scale*(lr*__fdividef((float)g_vals[j],sqrtf((float)s1_vals[j])+eps)); + break; + } + } } __syncthreads(); @@ -1469,11 +1475,14 @@ kOptimizerStatic8bit2StateBlockwise(T* p, T* __restrict__ const g, unsigned char { g_val = float(g_vals[j]); g_val *= gnorm_scale; - s1_vals[j] = smem_quantiles1[lane_id][c1s[j]]*absmax1[i/BLOCK_SIZE]; - s1_vals[j] = (s1_vals[j]*beta1) + (((1.0f-beta1)*g_val)); + if(!skip_zeros || (skip_zeros && g_vals[j] != (T)0.0)) + { + s1_vals[j] = smem_quantiles1[lane_id][c1s[j]]*absmax1[i/BLOCK_SIZE]; + s1_vals[j] = (s1_vals[j]*beta1) + (((1.0f-beta1)*g_val)); - s2_vals[j] = smem_quantiles2[lane_id][c2s[j]]*absmax2[i/BLOCK_SIZE]; - s2_vals[j] = (s2_vals[j]*beta2) + (((1.0f-beta2)*g_val*g_val)); + s2_vals[j] = smem_quantiles2[lane_id][c2s[j]]*absmax2[i/BLOCK_SIZE]; + s2_vals[j] = (s2_vals[j]*beta2) + (((1.0f-beta2)*g_val*g_val)); + } new_local_abs_max1 = fmaxf(new_local_abs_max1, fabsf(s1_vals[j])); new_local_abs_max2 = fmaxf(new_local_abs_max2, fabsf(s2_vals[j])); @@ -1509,9 +1518,12 @@ kOptimizerStatic8bit2StateBlockwise(T* p, T* __restrict__ const g, unsigned char # pragma unroll N_PER_TH for(unsigned int j = 0; j < N_PER_TH; j++) { - g_vals[j] = (T)(((float)g_vals[j]) + ((step_size*(__fdividef(s1_vals[j],(sqrtf(s2_vals[j])+(correction2*eps))))))); - if(weight_decay > 0.0f) - g_vals[j] = ((float)g_vals[j])*(1.0f-(lr*weight_decay)); + if(!skip_zeros || (skip_zeros && g_vals[j] != (T)0.0)) + { + g_vals[j] = (T)(((float)g_vals[j]) + ((step_size*(__fdividef(s1_vals[j],(sqrtf(s2_vals[j])+(correction2*eps))))))); + if(weight_decay > 0.0f) + g_vals[j] = ((float)g_vals[j])*(1.0f-(lr*weight_decay)); + } } // store: 0.85/1.44 -> 2.48/1.57 @@ -1623,23 +1635,26 @@ kOptimizerStatic8bit1StateBlockwise(T* p, T* __restrict__ const g, unsigned char { g_val = float(g_vals[j]); g_val *= gnorm_scale; - if(weight_decay > 0.0f) - g_val += ((float)p_vals[j])*weight_decay; - - s1_vals[j] = smem_quantiles1[lane_id][c1s[j]]*absmax1[i/BLOCK_SIZE]; - - switch(OPTIMIZER) - { - case MOMENTUM: - if(step == 1) - s1_vals[j] = g_val; - else - s1_vals[j] = (s1_vals[j]*beta1) + g_val; - break; - case RMSPROP: - s1_vals[j] = s1_vals[j]*beta1 + ((1.0f-beta1)*(g_val*g_val)); - break; - } + if(!skip_zeros || (skip_zeros && g_vals[j] != (T)0.0)) + { + if(weight_decay > 0.0f) + g_val += ((float)p_vals[j])*weight_decay; + + s1_vals[j] = smem_quantiles1[lane_id][c1s[j]]*absmax1[i/BLOCK_SIZE]; + + switch(OPTIMIZER) + { + case MOMENTUM: + if(step == 1) + s1_vals[j] = g_val; + else + s1_vals[j] = (s1_vals[j]*beta1) + g_val; + break; + case RMSPROP: + s1_vals[j] = s1_vals[j]*beta1 + ((1.0f-beta1)*(g_val*g_val)); + break; + } + } new_local_abs_max1 = fmaxf(new_local_abs_max1, fabsf(s1_vals[j])); } @@ -1662,16 +1677,19 @@ kOptimizerStatic8bit1StateBlockwise(T* p, T* __restrict__ const g, unsigned char # pragma unroll N_PER_TH for(unsigned int j = 0; j < N_PER_TH; j++) { - switch(OPTIMIZER) - { - case MOMENTUM: - p_vals[j] = ((float)p_vals[j]) - lr*(s1_vals[j]); - break; - case RMSPROP: - g_val = g_vals[j]; - p_vals[j] = ((float)p_vals[j]) - lr*(__fdividef(g_val, sqrtf(s1_vals[j])+eps)); - break; - } + if(!skip_zeros || (skip_zeros && g_vals[j] != (T)0.0)) + { + switch(OPTIMIZER) + { + case MOMENTUM: + p_vals[j] = ((float)p_vals[j]) - lr*(s1_vals[j]); + break; + case RMSPROP: + g_val = g_vals[j]; + p_vals[j] = ((float)p_vals[j]) - lr*(__fdividef(g_val, sqrtf(s1_vals[j])+eps)); + break; + } + } } // store: 0.85/1.44 -> 2.48/1.57 diff --git a/csrc/pythonInterface.c b/csrc/pythonInterface.c index 67bf2e5..7d5e654 100644 --- a/csrc/pythonInterface.c +++ b/csrc/pythonInterface.c @@ -110,7 +110,7 @@ extern "C" float eps, int step, float lr, \ float* quantiles1, float* quantiles2, \ float* max1, float* max2, float* new_max1, float* new_max2, \ - float weight_decay, float gnorm_scale, bool skip_zeros, int n) \ + float weight_decay, float gnorm_scale, int n) \ { \ name##_static_8bit_g##gbits(g, p, state1, state2, unorm, max_unorm, param_norm, beta1, beta2, eps, step, lr, \ quantiles1, quantiles2, max1, max2, new_max1, new_max2, weight_decay, gnorm_scale, n); \ |