ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

Add deepmind sample into bigdl-llm speculative decoding (#10041)

Browse source 
* migrate deepmind sample

* update

* meet comments

* fix style

* fix style
This commit is contained in:
Yina Chen 2024-02-01 09:57:02 +08:00 committed by GitHub
parent 72e67eedbb
commit 3ca03d4e97
1 changed files with 99 additions and 38 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

137
python/llm/src/bigdl/llm/transformers/speculative.py
View file

@ -80,31 +80,43 @@ def generate(
GenerationMixin.generate = generate
def sample(logits, return_probs: bool=False, do_sample: bool=False, top_k: int=50,
top_p: float=0.7, temperature: float=0.7):
def greedy(logits, return_probs: bool=False):
if return_probs:
all_probs = logits.softmax(-1)
if do_sample and top_k != 1 and top_p != 0.0 and temperature != 0.0:
_logits = top_k_top_p_filtering(logits.view(-1, logits.size(-1)) / temperature,
top_k=top_k, top_p=top_p)
output_ids = torch.multinomial(_logits.softmax(-1),
num_samples=1).view(logits.shape[:-1])
probs = torch.gather(all_probs, -1, output_ids.unsqueeze(-1)).squeeze(-1)
else:
probs, output_ids = torch.max(all_probs, dim=-1)
probs, output_ids = torch.max(all_probs, dim=-1)
return output_ids, probs
else:
if do_sample and top_k != 1 and top_p != 0.0 and temperature != 0.0:
_logits = top_k_top_p_filtering(logits.view(-1, logits.size(-1)) / temperature,
top_k=top_k, top_p=top_p)
output_ids = torch.multinomial(_logits.softmax(-1),
num_samples=1).view(logits.shape[:-1])
else:
output_ids = torch.argmax(logits, dim=-1)
output_ids = torch.argmax(logits, dim=-1)
return output_ids
def deepmind_sample(logits, return_probs: bool=False, top_k: int=50,
top_p: float=0.7, temperature: float=0.7):
prob_list = logits_to_probs(logits, top_k=top_k, top_p=top_p, temperature=temperature)
output_ids = multinomial_sample_one_no_sync(prob_list)
if return_probs:
all_probs = logits.softmax(-1)
probs = torch.gather(all_probs, -1, output_ids.unsqueeze(-1)).squeeze(-1)
return output_ids, prob_list, probs
else:
return output_ids, prob_list
def logits_to_probs(logits, top_k: int=50, top_p: float=0.7, temperature: float=0.7):
invalidInputError(top_k != 1 and top_p != 0.0 and temperature != 0.0,
"top_k != 1 and top_p != 0.0 and temperature != 0.0 if do_sample=True")
_logits = top_k_top_p_filtering(logits.view(-1, logits.size(-1)) / temperature,
top_k=top_k, top_p=top_p)
prob_list = _logits.softmax(-1)
return prob_list
def multinomial_sample_one_no_sync(probs_sort):
q = torch.empty_like(probs_sort).exponential_(1)
return torch.argmax(probs_sort / q, dim=-1, keepdim=True).to(dtype=torch.int64)
def clear_benchmarks(self):
self.first_token_time = 0
self.generate_time = []
@ -395,9 +407,13 @@ def speculative_generate(self,
logits = output['logits']
logits = logits[:, -1:]
logits[:, -1, :] = logits_processor(current_input_ids, logits[:, -1, :])
output_ids = sample(logits, do_sample=generation_config.do_sample,
top_k=generation_config.top_k, top_p=generation_config.top_p,
temperature=generation_config.temperature)
if generation_config.do_sample:
output_ids, prob_list = deepmind_sample(logits,
top_k=generation_config.top_k,
top_p=generation_config.top_p,
temperature=generation_config.temperature)
else:
output_ids = greedy(logits)
generate_ids[:, step] = output_ids
current_input_ids = output_ids
past_key_values = output['past_key_values']
@ -425,7 +441,11 @@ def speculative_generate(self,
else:
draft_past_key_values = past_key_values
draft_generate_ids[:, 0] = current_input_ids
draft_prob_list = []
tic = time.time()
random_probs = None
if generation_config.do_sample:
random_probs = torch.rand(max_step_draft, device=self.device, dtype=self.dtype)
# Draft model auto-regressively generate k tokens
# Early stop when prob less then th_stop_draft
for step_draft in range(max_step_draft):
@ -455,19 +475,25 @@ def speculative_generate(self,
logits = draft_output['logits']
logits[:, -1, :] = logits_processor(temp_input_ids,
draft_output['logits'][:, -1, :])
draft_output_ids, draft_output_probs = sample(
logits,
return_probs=True,
do_sample=generation_config.do_sample,
top_k=generation_config.top_k,
top_p=generation_config.top_p,
temperature=generation_config.temperature)
if generation_config.do_sample:
draft_output_ids, draft_probs, draft_output_probs = deepmind_sample(
logits,
return_probs=True,
top_k=generation_config.top_k,
top_p=generation_config.top_p,
temperature=generation_config.temperature)
draft_prob_list.append(draft_probs)
else:
draft_output_ids, draft_output_probs = greedy(
logits,
return_probs=True)
draft_generate_ids[:, step_draft+1] = draft_output_ids
draft_current_input_ids = draft_output_ids
draft_past_key_values = draft_output['past_key_values']
# check if draft prob is less then th_stop_draft
# Draft number + step >= max output token number
if draft_output_probs.item() < th_stop_draft or \
th_random = 1 if random_probs is None else random_probs[step_draft]
if (draft_output_probs.item() < th_stop_draft and th_random > 0.3) or \
step + step_draft + 2 >= max_new_tokens:
break
if self.device.type == 'xpu':
@ -533,21 +559,56 @@ def speculative_generate(self,
for i in range(logits.size(1)):
logits[:, i, :] = logits_processor(temp_input_ids[:, :input_ids.size(1)+step+i],
logits[:, i, :])
output_ids = sample(logits, do_sample=generation_config.do_sample,
top_k=generation_config.top_k, top_p=generation_config.top_p,
temperature=generation_config.temperature)
if generation_config.do_sample:
target_probs = logits_to_probs(logits,
top_k=generation_config.top_k,
top_p=generation_config.top_p,
temperature=generation_config.temperature)
else:
output_ids = greedy(logits)
if self.device.type == 'xpu':
torch.xpu.synchronize()
toc = time.time()
self.verify_time.append(toc - tic)
self.generate_time.append(self.draft_time[-1] + self.verify_time[-1])
# Compare drafts with target verified outputs
# Drafts start from [1, k]
# Verified output start from [0, k - 1]
# including the one generated by the base model
max_matched = ((output_ids[:, :-1] != drafted_input_ids[:, 1:]).cumsum(-1) == 0)
max_matched = max_matched.sum(-1).item() + 1
past_key_values = output['past_key_values']
if generation_config.do_sample:
draft_tokens = drafted_input_ids[:, 1:].squeeze(0)
draft_probs = torch.stack(draft_prob_list).squeeze((1, 2))
# q: target prob, p: draft prob
# q >= p: always accept draft token
# q < p: q/p prob to accept draft token
p = draft_probs[torch.arange(0, drafted_n_tokens), draft_tokens]
q = target_probs[torch.arange(0, drafted_n_tokens), draft_tokens]
accept_draft_prob = torch.minimum(torch.ones(()), q[:drafted_n_tokens] / p)
rejected_locations = (random_probs[:drafted_n_tokens] > accept_draft_prob).nonzero()
if rejected_locations.shape[0] == 0: # All draft tokens have been accepted
max_matched = drafted_n_tokens + 1
last_token = multinomial_sample_one_no_sync(target_probs[-1])
output_ids = torch.cat([draft_tokens, last_token])
else:
max_matched = rejected_locations[0].item()
p = draft_probs[max_matched]
q = target_probs[max_matched]
resample_prob = q - p
resample_prob = torch.where(resample_prob > 0, resample_prob, 0.0)
resample_prob = resample_prob / resample_prob.sum()
next_token = multinomial_sample_one_no_sync(resample_prob)
output_ids = torch.cat([draft_tokens[:max_matched], next_token])
max_matched += 1
output_ids = output_ids.unsqueeze(0)
else:
# Compare drafts with target verified outputs
# Drafts start from [1, k]
# Verified output start from [0, k - 1]
# including the one generated by the base model
max_matched = ((output_ids[:, :-1] != drafted_input_ids[:, 1:]).cumsum(-1) == 0)
max_matched = max_matched.sum(-1).item() + 1
max_of_max_matched = output_ids.size(1)
# Accept number is max_matched, min is 1
self.accept_num.append(max_matched)